First off, well done! This is incredible.
I think I know why the commenting data might not correlate well, however. Currently, the popular comment list has a time-dependent variable. The list is calculated by looking at the average score of commentors, but modified by a decay based on the time of their last comment. Due to this decay, the list is not sorted by the most highly rated commentor, but by the most highly rated recent commentors.
There are a few reasons why I do this. One is to allow for new users that make good comments to be noticed and appreciated. I also don't want the list to grow stale. Finally, I don't want those that comment frequently to be missed just because the average of their comments obscures a number of quality ones they have made. Thus, the list will always feature commentors that are appreciated, but not only those that have been most appreciated overall.
I was just discussing this with StephenBuckley earlier today, actually. I have put some thought to making the 'popular users' list a more dynamic as well. IMO the site is improved by a diversity of conversations, and at some level, I think it's counter-productive to reinforce the popularity of a small number of users.
That said, I can provide some time-independent comment data if you'd like. I think it would be interesting to see if badges do correlate to a user's average comment score, or perhaps their top or median comment score.
Really interesting analysis. I love this stuff.