A few of those I'll bet money weren't scanned. There were at least two Defiant CG models, one of which was infamously inaccurate around the nose. Likewise, the Miranda models weren't perfectly accurate (overly boxy, the impulse engines in the torpedo launcher), and the Nebula was built from parts of the Galaxy and lacked its distinctive puckered deflector and the missing window rows on the saucer.
Mojo posted alternate versions of some of the Starship Spotter wireframes to his blog a long time ago. While they're prettier than the ones in the book, they aren't quite as clear, and only a couple of the ships are ones that would've been "scanned" (and it's also possible those aren't the scanned versions, but they were rebuilt by the time the book was made) but you can get some sense of how they were structured.
For a quick example, so we're all on the same page, in my experience, 3D scanning is done by collecting points in space from the surface of the object. The model is then made by connecting each point with the points closest to it, resulting in a distinctive shattered glass or webbed structure, while hand-modeled objects have a much more regular structure in their mesh, either following simple geometric shapes, or following the contours of the surface. These are faces, not spaceships, but this
3D scanned bust and
this hand-modeled one should demonstrate what I'm talking about.
Now, to be fair, I don't really know what 3D scanning would've looked like in the '90s. It's possible they used one of those digitizing pens, and that got them nice, clean polyflow, or they did use a method more similar to the ones I'm familiar with today, which resulted in a tell-tale webbed pattern in the mesh, and then they used that model as reference and made the production model over it. Still, my instinct is the use of the term "scanned" is inexact and comes from explanations to people who weren't familiar with the technology using terms and analogies they knew better.