The phylogenetic inference of the gene trees is significantly hampered by the fact that we are pooling different MADS box class sequences in the same analysis. We can see this from the fact that we didn't achieve convergence in the mrbayes (and exabayes) analyses and that we are getting low aLRT in phyml. We can also see this from the fact that especially in the larger analysis (APETALA3 & PISTILLATA) very, very many (spurious) gene duplications, which throws the topology of the tree in doubt.

I have raised this issue on numerous occasions, but for some reason we have decided to plough on along this track.