There are plenty of theories that explain the data just as well as yours. Such as the voice modulation didn't mask their gender effectively (the demo videos are not very convincing), or women have different personalities to men which can be noticed in the types of things they say as much as the voice they say it in, or there's a selection bias in the types of women that use this site meaning the ones in the study are actually of a lower skill level but that women in general are not, or the same thing but with a selection bias for highly skilled men. Since there's a well known gender bias in the hiring practices in the tech industry, it's highly likely that there's a bias in the genders of people looking for work in the tech industry too. Since it's harder for women to get tech jobs, they're much less likely to quit a job on a whim.
Then there's the theory put forward in the blog post which was that women tend to become discouraged more quickly after one or two bad interviews where fewer men did. Once they excluded data from both men and women that only did one or two interviews the discrepancy went away.
Speaking of biases, if the first and only theory you come up with fits your own biases and strokes your ego it's very easy to stop there and smugly feel superior to half the human race, while being unaware of how wrong you are.