> E.g. how does this argument account for the data we have that someone with black sounding name will get less opportunity than someone with a white sounding name and an identical resume?
You're referring to a decades-old study that failed to replicate:
(This is extremely common in social sciences.)