I think it can depend on the population that you're comparing against. If the company is a law firm, the positions being compared are entry-level lawyers, and a requirement for the role is holding a law degree, then when evaluating the law firm's hiring practices you should probably compare against the population who hold the qualifications for the role (law degree holders).

If there are biases upstream of that point in the process (such as "those with a social security number ending in 9 are admitted to law school at one-sixth the expected rate"), then I'd expect an unbiased hiring process to result in that same discrepancy in those hired (presuming here that social security digits are entirely uncorrelated with performance in law school and the practice of law).

Should the law firm work to try to eliminate that upstream bias? Sure. Should the law firm work to hire in a biased way such that they end up with an evenly distributed outcome? I personally don't think so, but I understand others who think they should.