If I read this correctly, what you're saying is:
* Implicit discrimination is bad. I'm saying: To use that as a guide for your actions you need to have an implicit discrimination detector able to account for all the confounding variables.
* Explicit discrimination is also bad. I'm saying: As a corollary, there should not be explicit 'X need not apply' policies.
* Two wrongs make a right.