I personally find this fascinating--a biological basis carries with it a possibility of pharmacological treatment for those wanting it, and perhaps the design and validation of nonpharmacological approaches for those who don't or third parties seeking to reduce bias.
IIRC, there are a couple experiments in rat models along the lines of: put a male or a female rat in the same cage as an aggressive, territorial rat. The male rats are back to normal within hours-days. The female rats take days-weeks. IIRC some sort of neuronal transcription factor or something is thought involved.
Can track down a review if anyone's interested.
When a person decide to take up a job that is not typical for their gender and goes against ones gender identity, that decision might not bring the same confidence that someone taking a job where 80% of the people around are of the same gender as the person. That lower confidence get then translated into higher drop rates.
This alone might account for the difference in percentages there, assuming there's some hidden "adversity event" that caused the attrition.
Perhaps (speculating here), men handle rejection from jobs better because they handle rejection from the opposite sex better (due to years of honed experience) :O
It sounds sexist to say "grow a pair," but perhaps we need to teach women more confidence somehow.