That wiki page has a pretty substantial "Criticism" section. The Rosenberg et. al. paper even says their paper "should not be taken as evidence of our support of any particular concept of biological race." Their study shows that genetic distance is roughly linear with physical distance, barring obstacles like mountains and oceans. The consensus in the field, based on citations on that wiki page, seems to be that human diversity is a continuum, so any dividing lines you try to make to partition it will be largely arbitrary. It's like looking at a rainbow at trying to divide up the yellow from the orange from the red and so on: you can do that, but your color categories don't completely capture the spectrum visible.
You can make meaningful and consistent groupings in a continuum . For example you can make a distinction between poor and rich citizens, even tough the income distribution is a continuum. Same goes for the rainbow, the fact that it is continuous doesn't prevent people from being able to differentiate yellow and blue.
But even so the human genome is not continuously distributed, even if the difference is linear with distance, simply due to the fact that humans are not continuously distributed geographically. You can't handwave oceans and mountains away.
The main criticism of clustering is that the hierarchy of clusters that you obtain depend on the distribution of your samples, and that we currently do not have a good uniform sampling of human DNA. In short clustering can tell you reliably if someone is Irish or Italian, but may falsely identify the french as italian if there are not enough samples for french people.
