While I definitely think it might be a flawed sight I want to play devil's advocate and offer an alternative hypothesis.
So there is an issue with statistics for research, where if you collect a large enough sample size that is WAY more than what you need, you can use abuse of power (the large sample size that isn't needed) to make very small differences look like they are significant because the sample is sooo large. Similarly if you play enough games you will almost surely hit win streaks and loss streaks that are outside the normal range of what you go through generally. (It pretty much is a 100% chance that with a large enough sample you will get extreme outliers for length of win streak/loss streak)
Why does this matter? Because you can get promoted if your max MMR spike high enough on a lucky win streak, but you can only get demoted if you play a FULL tier below what you're listed. If that graph was MMR or ELO there would be no valid explanation for it because it can't occur that way given how an ELO system works. However, if you play enough games to brute force your way (play so many games a small statistical chance becomes very likely to occur at least once across the number you play) into hitting an outlier for win streaks you might be able to get promoted on a good win streak. You then will balance out your MMR on a loss streak eventually if you're playing the same way, but you're already in the tier above. Thus you end up with Gold tier account with Silver MMR because you played so many games you got lucky and promoted at some point. Then you can explain the graph because if everyone plays a ton of games compared to other regions then if the system is the same, the region with high play counts ends up pushing everyone (or at least a bunch of people who play a ton) into one tier above where their MMR is. Thus Silver is basically Bronze, Gold becomes Silver, Plat may just get bigger since Diamond is so small it may be hard to hit on luck, and Bronze disappears.
