The things that come to mind are that the min-hit on your Twin Snakes test is almost the same as the min-hit on your un-buffed test (If it was only your first unavoidable one, that's a hell of a coincidence, but it should be excluded from the sample nonetheless), and crits should either be removed from consideration or divided by 1.5.

Sample sizes appear a little small, too, but if I remember my napkin math correctly it should be sufficient for 99% confidence that it is one or the other.