Um. You are aware that humans from long-isolated groups can breed and that the majority of traits are polygenic (meaning there are lots of contributing alleles that interact with each other in frequencies that may or may not differ between two populations)? The range for many such possible traits in a human is wider than binary for polygenic traits but its edges are defined by a function of the average edges of the parents' inherited range (and so on), meaning that if you have two groups with very distinct frequencies and you mix 'em, the results will not be representative of the frequencies of either root source group, and in cases of colonization, that means interbreeding throws off the curves.