Probably a mistake on my part relying simply on Fiji. At least there, both V2 and V4 are shown as being 16bit but the value ranges are indeed very very different.

As this was more or less just meant as a trial to see if I can get QuickNII running with a 'custom' segmentation: do you think it can work with i.e. the data provided here: https://www.nitrc.org/projects/inia19?