They seem to have different teams working on different projects. On the one hand, we have the complete abortion that was the XL E-A and not long after we have the beautiful XL Reliant. Very much an inconsistent mixed bag. They’d better do a decent job aligning the paint for the STO version of the E-F. With all the high-contrast striping involved, that one won’t be as forgiving.
I don't even think they are different teams - they are just workers from another country, unfamiliar with Star Trek. That's why the models are so cheap (comparatively - considering the level of detail, paint, decals, ....)
The XL-A was pretty off. But it was obvious a lot of work went into it: The bridge dome is an
exact copy of the wrong bridge dome! That means they had extensive pictures of the model with the old bridge dome (promo-shots from before the movies were released), and really did a heck of a good job replicating that. They were just working off the wrong pictures - the bridge dome on the "real" movies had changed before filming.
In a stark contrast - for the Reliant there exists an official studio CGI-model. Converting that into a
mostly accurate model is easy. Hell, I'd wager all models are
mostly accurate! I LOvE their BoP models. But I honestly can't say if there are differences between te studio model, the CGI model and the Eaglemoss model - they look similar enough to me. Like a non-Trekie would never notice they accidentally put the wrong bridge on the Enterprise. That's only because for
us the hero-level get a much higher level of scrutiny.
The other issue is the side of the saucer, which,
yeah. That
was bad. But that seemed to be an issue often. The Kelvin-Ent also had a pretty bare saucer edge. I think the best approximation was the TMP-Enterprise (with it's three-layered structure) - also off, but worked, and the TOS Enterprise - which was also completel bare except windows. But that was a good fit for the model (and I think they put more effort in the original Enterprise than any other ship before - which they simply can't do for every model, otherwise they would work at a loss).
