John Hawks, who advocates the selection hypothesis has responded to the article with a blog post titled Selection, nuclear genetic variation, and mtDNA. This should highlight the controversial nature of the subject, and the fact that we are far from reaching a consensus.
Personally, I have always been somewhat skeptical of the idea that much can be learned about our most ancient past from genetic variation in modern populations. Modern populations are the result of so many evolutionary processes involving selection, drift, and intermixture, both in Paleolithic and most recent (and more tumultuous) Neolithic-historical times.
Genetic models seem to be fairly inadequate to capture the richness of human prehistory, as they more often than not tend to be fairly simple, with a lot of "best guesses" given to critical parameters. This has led to at least one palaeoanthropologist (Erik Trinkaus) proposing that we should do away with them altogether:
The analyses of extant human molecular data generally have little biologically relevant statistical power (whatever probability values their statistical computations may generate); most analyses use analytical algorithms whose biological assumptions and appropriateness are unstated, untested, and frequently untestable; many assume demographic stability over the past 50,000–200,000 years (see above); most consider the human populational dynamics of the past 30 millennia to have been trivial; many use distance statistics and graphic techniques (such as dendrograms), which deny the reticulate nature of human population evolution (hence assuming replacement); a number of them invoke molecular clocks whose reliability and precision within the time period of concern is undemonstrated and/or whose calibration (based on the fossil record) is simply wrong; and many employ living human samples of opportunity when those samples have biases relative
to the issue of modern human origins. And finally, all of them have a real-time depth of perhaps a century, and the interpretations based on those data are dependent on their analytical assumptions. This last point is evident in the large number of articles concerned more with the analytical techniques and their assumptions than with interpreting the available data.
It would be heartening to think that things are better in palaeoanthropology, but non-genetic factors affecting morphology, significant dating uncertainties, unavailability of key fossils for study, as well as the largely spotty record (a few skulls over tens of thousands of years) probably make that field as unreliable as genetics.
Perhaps we don't yet have either the data or the tools to make authoritative inferences about our origins quite yet, but the process is certainly interesting!