I’m stupid. It helps to use the correct cover fractions when plotting global totals. The pre-industrial line was using the cover fractions from my no crop run, so they were too high also. Correcting that brings them very close to each other.
Nevertheless, it remains that the existing update_cable_vegfrac.py is not quite suitable for simulations with large land-use change. My version of it that fixes the above issues is here, in case anyone is interested in the future: