Phylogeny of Cattleyas

Orchid systematics/taxonomy has been unstable over the last few years as old hypotheses on evolutionary relationships based on morphology have been re-examined using molecular data.  Both kinds of data can either provide poorly supported answers or misleading answers about which species are most closely related to which other species.  This has led to all kinds of naming wars as botanists propose new, sometimes conflicting evolutionary trees, often based on small datasets, that upset horticulturists because they are asked by some organizations in the orchid world to change the names of their plants.  And then new trees are proposed and orchid growers are asked to change their plant names again. And again.  Here I show one of the newest evolutionary trees for Cattleya from a 2009 paper by van den Berg published in Annals of Botany, addressing the phylogeny of the Laelinae tribe, which includes Cattleya, Epidendrum, Encyclia, and several other neotropical genera. 
Here are a couple of salient points.  First, some of the surprising relationships illustrated here are poorly supported.  For example, although Brassavola is a well supported group, the relationship of Brassavola to Cattleya is ambiguous.  Second, the genus Guarianthe (which contains what used to be called Cattelya skinneri, C. aurantiaca, and C. bowringiana), makes no sense from an evolutionary perspective.  Third, the genus Rhyncholaelia (which contains what used to be called Brassavola digbyana and Brassavola glauca) also makes no sense.   These two species are clearly not with the other Brassavola, but they will probably end up in Cattleya, if you asked me to bet.  All the arguments about naming in these species would vanish instantly and forever if all the Cattleyas, Brassavola, Guarianthe, and Rhyncolaelia were just called Cattleya.  That probably won't happen, but another reasonable alternative will probably to just have two genera here, Brassavola and Cattleya - additional data will be required to see if this is legitimate. Perhaps I'll talk about some other conclusions from this paper later.