Let me try to understand the situation when looking at the accused person:
1. There is someone who solves lots of puzzles.
2. He is rating all those puzzles.
(Maybe 3. He usually rates puzzles lower than the average rating of the puzzle.)
(Maybe 4. The given ratings are randomly chosen.)
(Maybe 5. The given ratings are not randomly chosen, but biased on some parameters. Like 'solveability', 'niceness', 'puzzle type', 'publishing frequency', 'solution availability', 'current rating of the puzzle', 'time of the day', 'mood', ...)
Some consequences of those points:
1. Lots of puzzles get one solver more.
2. Lots of puzzles are getting the actual rating value faster. (Which is good, as the average amount of solvers per puzzle seems to have declined.)
(Maybe 3. The skewed distribution of ratings gets pushed a bit in the right direction, making rating differences bigger, and thus making ratings better comparable, and thus the rating system more useful.)
(Maybe 4. Some puzzles will get lucky, some puzzles will get unlucky.)
(Maybe 5. The puzzles are rated as they are meant to be rated.)
Some mathematics:
6. When someone gives a rating which is lower than the current average rating, then the average rating goes down.
7. When a puzzle has a rating of more than 75%, any person that is not giving 5 stars is lowering the rating.
8. To achieve a rating of 90%, most people need to give 5 stars.
9. If all people are only using 4 and 5 stars to rate puzzles, then there are only 26 different possible ratings (75% to 100%).
10. If all people are using all 1 to 5 stars to rate puzzles, then there are 101 different possible ratings (0% to 100%).
11. The impact of a given rating is bigger when using 1-5 stars than when using only 4 and 5 stars.
12. It's easier to distinguish values on a scale that has lots of different values than on a scale that has only a few values. (E.g. if puzzle A has 100 top-votes and 0 bottom-votes, and puzzle B has 99 top-votes and 1 bottom-vote, then when only using 4-5 stars both have 100%, but when using 1-5 stars puzzle A has 100% and puzzle B has 99%.)
Some psychology:
13. Some people always blindly give 5 stars, just because.
14. Some people always blindly give 1 star, just because.
15. Some people vary their ratings between 4 and 5 stars, depending on how they liked the puzzle.
16. Some people vary their ratings between 1 and 5 stars, depending on how they liked the puzzle.
17. For some people, a 4-star rated puzzle is bad.
18. For some people, a 4-star rated puzzle is good.
Putting it together:
If puzzles have a rating of 90% or more, only a few people are giving bad ratings.
Chances are, that most of the people are always giving 5 stars, no matter what.
Which in return means, there are only a few people who regularly also give less than 5 stars.
If there is one such person who is regularly also giving less than 5 stars, and who is also solving a lot of puzzles, and if you are a puzzle author that is publishing a lot of puzzles, then chances are, that this same person solves many of your puzzles. And then obviously this one person is usually the one that gives lower-than-average ratings.
If you "catch" someone giving you a low rating, it feels bad. But in this very situation, he is doing the right thing and you are not: He uses the rating system in a way that he thinks is correct. You are focusing on one bad vote instead of looking at the overall average. Tastes are different. Perceptions are different. People are different. You don't have to please everyone. Yes, the rating system is broken: it should be anonymous, so nobody gets himself into this psychological problem of being able to connect votes with users. But here we are. I know it's hard to not constantly look at those ratings.
Maybe something to think about: Imagine there is a political election going on. You are the person who has to count ALL those millions of votes. Imagine your emotions after the first opened vote: Either your are totally happy because the right person is winning, or you are completely devastated because it's the (in your opinion) wrong person. The second vote might or might not change everything. So might the third vote, and so on. Chances are, that at the beginning you will see a lot of ups and downs. A lot of "good" votes that correlate with your opinion. A lot of "bad" votes that anti-correlate with your opinion. Sometimes the "good" ones are winning, sometimes the "bad" ones are winning. At the end, there will be just an average, and one single vote won't count much.
Now, imagine you do the same thing, but you can also read the names of the voters. At the beginning, you might be angrily thinking something like: "Damn, this person again, always voting for the wrong ones.". But at the end, you might be cheerfully thinking something like: "Haha, this person again, always voting for the wrong ones.".
In both scenarios, looking at it in the beginning gives you pain. Just avoid it. The difference is, in the second scenario you also develop some connection to the first people that voted.
1. There is someone who solves lots of puzzles.
2. He is rating all those puzzles.
(Maybe 3. He usually rates puzzles lower than the average rating of the puzzle.)
(Maybe 4. The given ratings are randomly chosen.)
(Maybe 5. The given ratings are not randomly chosen, but biased on some parameters. Like 'solveability', 'niceness', 'puzzle type', 'publishing frequency', 'solution availability', 'current rating of the puzzle', 'time of the day', 'mood', ...)
Some consequences of those points:
1. Lots of puzzles get one solver more.
2. Lots of puzzles are getting the actual rating value faster. (Which is good, as the average amount of solvers per puzzle seems to have declined.)
(Maybe 3. The skewed distribution of ratings gets pushed a bit in the right direction, making rating differences bigger, and thus making ratings better comparable, and thus the rating system more useful.)
(Maybe 4. Some puzzles will get lucky, some puzzles will get unlucky.)
(Maybe 5. The puzzles are rated as they are meant to be rated.)
Some mathematics:
6. When someone gives a rating which is lower than the current average rating, then the average rating goes down.
7. When a puzzle has a rating of more than 75%, any person that is not giving 5 stars is lowering the rating.
8. To achieve a rating of 90%, most people need to give 5 stars.
9. If all people are only using 4 and 5 stars to rate puzzles, then there are only 26 different possible ratings (75% to 100%).
10. If all people are using all 1 to 5 stars to rate puzzles, then there are 101 different possible ratings (0% to 100%).
11. The impact of a given rating is bigger when using 1-5 stars than when using only 4 and 5 stars.
12. It's easier to distinguish values on a scale that has lots of different values than on a scale that has only a few values. (E.g. if puzzle A has 100 top-votes and 0 bottom-votes, and puzzle B has 99 top-votes and 1 bottom-vote, then when only using 4-5 stars both have 100%, but when using 1-5 stars puzzle A has 100% and puzzle B has 99%.)
Some psychology:
13. Some people always blindly give 5 stars, just because.
14. Some people always blindly give 1 star, just because.
15. Some people vary their ratings between 4 and 5 stars, depending on how they liked the puzzle.
16. Some people vary their ratings between 1 and 5 stars, depending on how they liked the puzzle.
17. For some people, a 4-star rated puzzle is bad.
18. For some people, a 4-star rated puzzle is good.
Putting it together:
If puzzles have a rating of 90% or more, only a few people are giving bad ratings.
Chances are, that most of the people are always giving 5 stars, no matter what.
Which in return means, there are only a few people who regularly also give less than 5 stars.
If there is one such person who is regularly also giving less than 5 stars, and who is also solving a lot of puzzles, and if you are a puzzle author that is publishing a lot of puzzles, then chances are, that this same person solves many of your puzzles. And then obviously this one person is usually the one that gives lower-than-average ratings.
If you "catch" someone giving you a low rating, it feels bad. But in this very situation, he is doing the right thing and you are not: He uses the rating system in a way that he thinks is correct. You are focusing on one bad vote instead of looking at the overall average. Tastes are different. Perceptions are different. People are different. You don't have to please everyone. Yes, the rating system is broken: it should be anonymous, so nobody gets himself into this psychological problem of being able to connect votes with users. But here we are. I know it's hard to not constantly look at those ratings.
Maybe something to think about: Imagine there is a political election going on. You are the person who has to count ALL those millions of votes. Imagine your emotions after the first opened vote: Either your are totally happy because the right person is winning, or you are completely devastated because it's the (in your opinion) wrong person. The second vote might or might not change everything. So might the third vote, and so on. Chances are, that at the beginning you will see a lot of ups and downs. A lot of "good" votes that correlate with your opinion. A lot of "bad" votes that anti-correlate with your opinion. Sometimes the "good" ones are winning, sometimes the "bad" ones are winning. At the end, there will be just an average, and one single vote won't count much.
Now, imagine you do the same thing, but you can also read the names of the voters. At the beginning, you might be angrily thinking something like: "Damn, this person again, always voting for the wrong ones.". But at the end, you might be cheerfully thinking something like: "Haha, this person again, always voting for the wrong ones.".
In both scenarios, looking at it in the beginning gives you pain. Just avoid it. The difference is, in the second scenario you also develop some connection to the first people that voted.