Interesting and somehow contradictory. Your workflow delivers lots of pixels with a rather limited potential for tonal finesse, hmmm.
What I do is similar, but I simply take one shot only. This is for cataloging and normal use. In case I want something really big, I can always rescan and stitch and live with that huge file, it’s not 100 in 100 but maybe just 1 in 100 - or not even that after all.
How big a difference of keeping four shots vs. one stitched DNG that contains those four shots? Unless you keep both original and stitched files, the difference might be bearable.
The magic lies in keeping track of the negatives. There’s at least one post dealing with this: Organizing Negatives For Scanning