To your point, a bias in the encounter rate (assuming it exists) is a significant problem, and one that would need to be addressed. But it's also a distinct problem, and there's no reason to believe that the violence problem and the disproportionate-encounter problem would be amenable to the same solution.