Seriously, no, Top 100 parses is basically completely useless and misleading for class balance discussions. I wrote a blog post about this waaay back when the log collation site in use was stateofdps.com;
How not to use stateofdps.com and worldoflogs.com. The specific numbers used as examples are out of date, but it's a post about statistical theory and those numbers are just examples; the same facts hold true today. There's a reason everyone stopped using stateofdps.com and started using raidbots.com (which serendipitously started out around the same time as I posted that; the devs there contributed to the thread in fact). It's because "top parse" models are bad for class balance discussion; it biases things towards who's had the best luck doing gimmicky stunts, or who's got the greater RNG and got a lucky parse to upload, rather than class performance and player skill.
If you want to control for player skill, just look at (H) content; those players are performing well enough to get (H) modes down. Trying to go by top 100 brings in far too many irrelevant factors that cannot be controlled for, factors you CAN control for simply by using "all parses".