My investigations continue. I've gathered data on a lot more stars, broadening the search to many varieties of G, K and M stars.
I'll do a fuller write-up at the weekend I expect.
In the meantime, a few things stand out:
VZ stars are almost certainly low-metallicity stars (which makes sense and is as we suspected); the ones I've looked at are considerably lower effective temperature than an equivalent ordinary star of the same spectral class.
Spectral classification being a black art, it's partly done based on the relative strengths of metal and hydrogen lines, meaning a star with less metals in the first place appears to be a different (higher, in practice) class than it would normally be.
There are weird periodicities in the frequency of generated stars. Some types (e.g. G7VAB) are far more common than others (e.g. G8VAB.)
Plain "V" stars (e.g. G5V), which so far as I can tell are ones which have been hand-placed, can sometimes fall into the ranges you would expect from auto-generated stars, and can sometimes be both outside those bounds and outside (so far as I can tell) normal bounds of classification, for instance I've found a G1V with an effective temperature of 7205K. Toasty.
Almost all of the stars I've looked at have habitable zones which are easily predicted based on their temperature and radius. In at least one instance, though, (HIP 14383, which has unfortunately been first scanned by a CMDR who the game currently won't let me report for obscenity) the planets in the system appear to have had their temperatures calculated as if the star was much brighter (possibly using negative albedo - that fits to within a few degrees) - they are considerably hotter, even the one without an atmosphere, than a blackbody would be at the same distance. Bizarre bug. Any thoughts?