It depends on how they arrive at their estimate: If they just assume some large number of black people will vote because that gives them the result they want, then yeah that's wrong. And they wouldn't be taken seriously as a pollster if they did that.

But if they're basing it on the number of black poll respondents that report that they're going to vote/have already voted, then that's perfectly reasonable and you can't argue that 'no way polls are rigged cause Clinton isn't black like Obama'.