It turns out that the problem is a bit more interesting than I thought, and I'm inclined to fix it by changing how the regular Gen2 ci_hsc works a bit.
The issue is that the brightObjectMasks are in the repo at $CI_HSC_DIR/DATA, and they're defined with a (tract, patch) data ID. But there's actually no skymap defined in that repo - the skymap is added to the $CI_HSC_DIR/DATA/rerun/ci_hsc repo. So the gen2convert is quite reasonably warning that it can't identify the skymap that defines the tract and patch data ID for brightObjectMask, and then skipping them.
This could be fixed by either defining the skymap directly in the root repo or putting the brightObjectMasks into the rerun. I'm leaning towards the latter; please shout if you have an opinion.
We should also keep this in mind when it comes time to run gen2convert on the real repos on lsst-dev or tiger; I bet they suffer from the same problem.
You may be interested in the pattern Jim Bosch used in DM-17048 for doing something similar with ref_cats.