plot_age_posterior_distributions.R

This seems to be the case with the COI gene that tends to pull your timings towards the past. Saturation in the 3rd codon position might be accused for responsibility (you can inspect the saturation level with some cool plots).

Brandley et al (2011) use a nice way to figure out whether some of your genes might be giving very different time estimates for your trees. You can select 1000 random trees from your BEAST run and plot the distributions of the ages for the crown group of different genes, different codon positions and the combined analyses. Something like this plot consisting on a simulation of a gen1 estimating a crown age of 30Mya, gen2 estimating an age of 50Mya and the combined analysis giving an age of 40Mya.