I appreciate how much effort you've put trying to work this out by pure theory. But some of it, in practice, is just flat out wrong (pun intended).
Corrections I'd offer based on real world experience of designing and deploying such systems to institutions who then rigorously test them and refuse to pay if you don't meet objective measurements:
- A Drum Scanner no longer produces the best results; we've surpassed that legacy standard bearer.
- Wet mounting is no longer preferred; in addition to the productivity and conservation concerns the fundamental advantages are not what they were with the underlying technology used in a drum scanner. When inbound illumination is direct rather than angled and diffuse rather than collated the topography of dust and scratches is not nearly as problematic. Instead our system uses a proprietary high-resolution (fine grain) anti-newton ring glass that has a far higher resolution than the ANR glass of the film era. This ensures complete flatness without the downsides of wet mounting or legacy low-resolution ANR glass.
You are absolutely right though about flare, and planarity/alignment. We manufacturer our system in the USA using metal cut and tested to extremely tight tolerances, and then use a laser system for the final alignment calibration of the system. Our carriers are likewise made of metal (solid plate metal, not flexible sheets) and each is carefully checked for flatness. Even a slight misalignment causes measurable loss of quality across the frame. In fact though that is not a problem limited to camera-based scanners; we've had several clients whose legacy film scanners exhibited intra-frame sharpness loss, including two separate clients who sent their film scanners in for pro service only to wait several months to have them returned with the same issue (along new axis). If anyone is interested in testing their scanners homogeneity of resolution, I suggest an
ISA target along with Golden Thread software or the free variant of that software produced by the Library of Congress. Just be prepared to be disappointed; I've yet to see a single legacy scanner test that didn't exhibit some measurable misalignment.
You also missed out on several other factors such as focusing accuracy, external vibration mitigation, camera-generated vibration mitigation, focus stability on cameras designed to point outward rather than downward, flatness of field (even on repro lenses; there are no 100.00% flat-field lenses, only lenses designed with varying levels of rigor on this spec), thermal output of lighting, the spectral quality of the lighting and the generation of a matching profile, uniformity of illumination, the metameric error from specific emulsions, dust mitigation, eliminating carrier-based bend or displacement during handling, and assuming the volume of scanning is high the topics of automatic cropping, film negative tone inversion, multiple-derivative output, and naming / file organization. I'm probably missing a few.
Of course, how much any of the above raised issues matters to you depends on what level of quality you want/need to achieve and what volume you're scanning. If you just want to make some pretty 8x10s or 11x14s of a few dozen (or even a few hundred) slides then you don't need anything fancy to have a great experience. If you're looking at thousands or tens/hundreds of thousands of pieces of film from which you want to make archival-grade never-scan-again files that would beat a drum scanner, then, frankly, you do.
If I come off sounding a bit drum-beating about our system its because we've spend the last several years systematically solving these problems and I'm darn proud of what we've accomplished in doing so.