The issue 'race is correlated with X' is not materially affected by how exactly race (especially its borders) is defined and taxonomical problems - it's important in it's own regard, but in this regard, it's not the core problem, it's a red herring distracting from the core question of "does race affect X'; and that question is the same, with the same answers, no matter if in your analysis you choose to model race as discrete borders or as a continuum.