This data is interesting, but even if you ignore the weirdness introduced by gender localization and small sample sizes, it doesn't compare directly to the gender numbers that came out of the editor surveys.  I don't have actual statistics on it but going from memory I'm almost positive that the software settings indicate a consistently higher proportion of women than men than direct surveys like editor trends do. This tool can't provide enWP stats and I don't know them offhand, you can look at other en projects to get an idea of how different the numbers reported this way are from the numbers reported directly by survey.  This tool has enWikiversity reporting ~20% women, enWikibooks ~16%, and enWikisource ~17% - all significantly higher %'s than have been found using other methods.  So, instead of ptWikiversity reporting 21% women vs the 8.5% you would expect enWikiversity to have going off of this year's survey results, it would be ~21% vs ~20%.  

This is certainly something worth exploring and I am certain that there are language community based differences in gender ratios, but it is good to be aware of the limitations of the dataset.