Understood. I'm envious of your data set. I did some analysis of the kind you're talking about on my own (with a lower quality set of numbers than yours). The analysis part was fun. Pulling together the data was a pain. Your results so far look consistent with mine, but you're also looking at lots of issues I couldn't begin to explore with my limited data set.

Have you played at all with using something other than than a horse's top as the anchor for your analysis? I'm no statistics expert, but most of what I've read on the subject suggests using median or mean (rather than high or low) to establish the "location" of a distribution. In my own dabblings, I found a horse's median figure to be more useful than its top or mean figure.