By failing to account for obvious and known confounds, the graph is rendered significantly less meaningful.
By way of analogy, one could look at rainfall averages over the whole of the US and conclude that some place is/is not in a drought. You could do this by ignoring the obvious confound that different places get different amounts of rainfall. For some places, it would be true. For others, not.
That's the crux of the matter.
Most people unhappy about the 'pay gap' are indeed implying that employers are paying double (or 21% more) wage for a y chromosome.
If the wage gap rather has to do with decisions employees make themselves (such as who stays home with the kids and thus works fewer hours / has less experience / doesn't get promoted) - then it's not really the problem of employers to contend with.
I don't undestand your point. Those are exactly the mechanisms most likely responsible for the pay gap. No employer would ever pay double wage for an y chromosome, but it's not unlikely that preconceived gender roles influence decisions about who gets promoted, who stays home with the kids and stuff like that.