I don't know. If I'm reading into it correctly, the stereotype is easier to base off of due to familiarity.
For example, you can have a very masculine female (or basically someone you wouldn't want to piss off that might also be able to out-drink you in a drinking contest (Male > Female, by a margin of 50-75), and you can have rather effeminate (straight) males that are "Metro-sexual" breeds or something of the sort (same margin, swap points). As far as I can tell, based on your familiarity of said stereotypes, it can or can't be accurate. Minor bias as to how much of said stereotype you understand or have observed.
I, of course, state the potential bias due to the fact that one could be around very masculine women and account that as a norm of feminine behaviors, and same with the masculine like I stated before for a majority of their lives, and thus bias their results due to their familiarity being different than the majority norm. But all that aside, I think they're counting on familiarity of said stereotypes based on the majority (Feminine = Barbie, Masculine = Hot Wheels, so to put it) as the basis of the norms being used.
EDIT:
Basically, at least from what I've observed, even with my own stats (see previous post for how I analyzed it), I think I did a fair assessment of what mine meant. Of course, my margins are on a high-level (near-maxed out on masculine, and halfway feminine). If you lack male and female stereotypes, you could be a spambot, though that could be a stereotype of autonomous beings, and be offensive to Skynet.