Well, it turns out that I was substantially correct, and the other day a fix went in which removes from the database the old useless data which was inflating the results. The code arrived on site in yesterday's code push, and so today's newly generated statistics are possibly the first accurate ones for several years. (This follows the fixing of the supportrank stuff in mid-October after well over a year of it not working.)
Here are the gender stats from yesterday:
- Male: 2500541 (32.6%)
- Female: 5169632 (67.4%)
- Unspecified: 2558249
And here's what it says today:
- Male: 1256235 (32.4%)
- Female: 2620778 (67.6%)
- Unspecified: 1418850
Quite some difference! Although it's interesting to note that the percentages have hardly changed.