The article is well done and worthy of being published.
However, I have some doubts about how the analysis of divergence times was conducted. Is it correct to calibrate the tree outside the considered ingroup since Beast returns local clocks? If the rates along the outgroup lineages differ greatly from those of the ingroup, your results may be skewed.
Furthermore, the authors compare their estimates with the previous ones obtained by calibrating only with geological events but do not consider 95% HPD at all. Considering the intervals, the estimates are not that different at least regarding one of the two limits.
I think that divergence times estimates are strictly dependent on the considered calibration node, and I suggest considering these concepts in the discussion.