Here's my thinking:

First, consider that the sample size vs population size of CEOs is goddamn tremendous, and the size of both pools is pretty substantial. It's one thing to try to select a stratified sample of 10 out of 20; it's another to select a stratified sample 500 out of millions. I would expect the transition point to be around when people stop being people, and start being statistics.

Second - It's like dark matter, and then it gets more complicated. Notionally, men and women are equal in capability; given just that, I'd expect to see a sample representative of the population demographics. This is not what is observed, so we can posit a) inequality in capability and/or b) other forces acting on the system.

A isn't born out by other experiments, leaving B. So then it becomes a question of: What are those other forces AND are they ones we want to keep?

Contrast with age, which clearly has a correlation with inequality in capability, so less of the disparity with the theoretical stratified sample must be explained by B.

Of course, shit gets complicated fast; the correlation of skill with age is not only non-linear, but is more like a non-linear probability function of skill given an age; self-selection might at first seem to be a reasonable B force, but then you have to consider what's causing the self-selection, and what effect that should have on the "equally capable" population sizes...

At 55, it looks like CEOs should be equally split down the middle. That this is not the case means there's more forces in play than explained by the "theory".