I agree that there are many plausible explanations for why code from women would be rejected at a higher rate at Facebook. I don't think that demonstrates bad stats though. I don't think your collaboration suggestion is very likely (given other public large scale analysis of male/female PR acceptance rates that showed them to be roughly equal) but I don't think that undermines your overall point.
I don't think it's material whether it is overt or subconscious bias leading to these results - either one would be a big problem that Facebook should recognize and work on.
As an example, I'm sure of you're heard of the 77-79 cent wage gap. I'd call that bad stats -- what the study was actually comparing was the average income of the entire male US population and the entire female population. When you correct for job title the gap is actually closer to 95 cents on the dollar -- and you're still not correcting for other factors, such as the fact that men work more hours than women and tend to be more aggressive negotiators. But people read the study, took the data on face value, and now the wage gap is a cultural meme we can't get rid of.
I think the same thing applies here -- a slice of the data looks funny so people cry sexism. I wouldn't at all be surprised if the difference goes away if you adjust for level, experience, technology used, product, background, etc, etc, etc. Publishing stats like that is dangerous unless you also release the data set and the method you used to derive your stats.
Why I wouldn't take it at face value: rejecting CRs sucks. Every round of review takes time away from the reviewer as well as the reviewee. Unconscious biases exist, sure, but it's hard to argue that people are actively making their life more difficult because of them.