I haven’t experimented with it yet (I think it was you who recommended this setup to me Gotz) that for the shots of the bottom, if you have a completely different background from your initial shots, Reality Capture will end up correctly aligning the top and bottom. In the example I saw, it was using a sheet of newspaper for the background on all the shots except that the group shot for the bottom had a different sheet of newspaper.
The examples I saw with this were of a shoe, where it was flat on a turn table for the majority of the shot then a round or two shot with the shoe flipped on its side (not completely upside down) that way there was enough overlap between both groups of images for RC recognize.