Scanning the study now--I guess the answer is probably that it's not 100% consistent.

Quote
Performance deficits triggered by stereotype
threat were more likely to appear when the young
female sample was playing males in more challenging situations. Specifically, females did worse
than expected when competing against a strong
opponent (achieving 56% of what would be predicted based on prior ratings), a moderate one
(82% of expected performance), an opponent
from a higher grade (73% of expected performance), and an opponent of the same grade(83% of expected performance). This supports the contention that anxiety, arousal, and other
processes related to stereotype threat only impair performance on difficult, challenging tasks...When competing
against weaker opponents and opponents from
lower grades, young females performed about as
expected or only marginally worse so. This suggests that playing someone perceived as weaker
or younger may buffer the negative impact of
playing against a male chess player.

And, of course, girls don't always play boys!-- though I don't know what percentage of players at tournaments ARE girls.