Or: discrimination is baked into the assignment of job titles.
Right. How would we tease that apart? When two people have the same years of experience in the same job title, but then one of them gets a new title and the other one doesn't-- what is that? Just from the data, we won't really know whether it's discrimination or difference in ability.
It'd be cool if we could have objective measurements of productivity. But if we could have objective measures of productivity, I wouldn't be wasting my time with macro data analysis, I would be selling the system to companies and trying to get rid of subjective employee reviews.
_Nevertheless_ the numbers are getting better as you control for things! It suggests that, as you control for more factors, the wage gap will shrink to nothing.