I'm sure I did mean to imply that it was an artificial construct, yes. Gender expression and identity are more continuous than people tend to believe.
For the interested, this seems like a pretty good overview of gender identity (primarily from a developmental perspective, of course):
http://steinhardt.nyu.edu/appsych/opus/issues/2011/spring/ge...
If the gender data is somehow critical to the data then it's worth being accurate. If not, why even ask?
Anyway back on topic:
Open field text boxes come with a practical cost of being much more difficult to aggregate, and let's be honest, the gender data is probably not important enough for GitHub to expend too much energy on. The more costly you make the data to mine, the less likely that the data will be used effectively.