Three-stage colonization model for Americas reconsidered

I had blogged about this model before. From the earlier paper:

Bayesian skyline plots, which provide dynamic representations of population size changes over time, indicate that Amerinds went through two stages of growth ≈40,000 and ≈15,000 years ago separated by a long period of population stability.

This model was now criticized by Fagundes et al. who discovered that the early stage of the model was an artifact of the inclusion of mtDNA sequences of non-American origin:

Our results refute the specific details of the “three-stage model”, since the early stage of expansion into Beringia followed by a long period of stasis could not be reproduced in any mtDNA data set cleaned from non-Native American haplotypes.

This is acknowledged by the authors of the three-stage model in Mulligan et al:

We re-assess support for our three stage model for the peopling of the Americas in light of a recent report that identified nine non-Native American mitochondrial genome sequences that should not have been included in our initial analysis. Removal of these sequences results in the elimination of an early (i.e. ~40,000 years ago) expansion signal we had proposed for the proto-Amerind population.

The authors do however claim that:

Thus, our three stage model remains an important and useful working hypothesis for researchers interested in the peopling of the Americas and the processes of colonization.

What's the lesson from all this? Many of the conclusions drawn by geneticists are very sensitive to data outliers: 9 out of 77 sequences that shouldn't have been included was enough to provide support for the 40kya stage of growth, which was now revealed to be spurious.