I do not know if I am right or wrong. Neither does anyone else who does not work at Blizzard.
For clarity, I am not arguing that I am right or wrong, I am simply saying that there are some peculiarities in the presented data that seem counterintuitive. Other posters speaking in absolutes “MicroRNA was wrong” also do not know the merits of that statement given what you just wrote.
Yes and Blizzard also mentioned that the leaderboard data overestimates the “average” player. Therefore, you can see how their numbers compare.
Complete data is not needed to analyze relative build strength. What is easily available in bulk for the leaderboard are GR tier, paragon level, clear speed. Manually, you can get the legendary gem levels, augments, and builds for the leaderboard clears in game.
I have actually done this in the past and posted on it. One oddity at that time was I previously assumed that the legendary gem levels for a “meta class” would be higher than a non-meta class (excluding capped gems and pain enhancer). It turns out that wasn’t the case for the “average” players I analyzed at the top end of the leaderboard.
Moreover complete data actually is problematic. You can think of it this way: Informative, high quality data informs, bad/irrelevant data muddies the waters.
For example, there could be a high paragon players that actually avoid the leaderboard to avoid “detection”. This would make “complete data” for classes that are easily botted skewed.