[anigbrowl]

But, er, doesn't the observation that the actual ratio is different evidence constitute "evidence"?

No. That tells you there's a difference, but absolutely nothing about the causes. You may not know of an artificial causal effect that inhibits greater participation in software by women, but not knowing about it doesn't mean that it doesn't exist (unless you have falsified all available hypotheses, naturally).