Hm, I missed that link to the chart before, and I was trying to respond to the "cognitive diversity -> effectiveness" discussion. You're right: it's hard to see the chart supporting my claim. (There are some hints at a sweet spot there in the middle, except for that drop at 50/50!)

Link to chart: http://hbr.org/2011/06/defend-your-research-what-makes-a-tea...

The variance, just from glancing at that chart seems extraordinarily high. I'd be very hesitant to draw conclusions.

Factoring in the variance, it looks to me like groups that had an even split did the best. Groups predominantely of one gender suffered.

The "women are better" trend line may partially be accounted for from the fact that there were simply more groups with more men.

Could be that in groups without women, men underperform because of a lack of stimulation.