So if companies are rejecting someone that means as whole (read net), they are expecting more bang for the buck.
Unless your model accounts for such things, it's too simplified. Particularly when it comes to human factors. And if your model is that deficient, it's not really providing useful insights.
Modeling complex phenomena well is tough. But you need to have enough humility to understand when your toy model might be deceiving you.
If the shoe fits....
At least that's what I was told at the talk about diversity and inclusion during my orientation at one of those big tech companies.
Apparently it's because we benefited from being able to make quick judgments before civilized society was a thing.
If there is a strong shortage of supply, you can be sure a lot of people will get over their prejudice -- but in labor markets, there often tend to be more offer than demand, so companies can choose whoever they want, without hurting the business whatsoever.