I was trying to swap out the manual conversions in favour of relying on astropy, but unfortunately I run into multiple issues with the magnitude errors as well as the plotting.

Spending some time with debugging I sense that it's a mixture of missing features (e.g. handling masked data with nothing masked, to potential ufunc issues around magnitude units and errorbar plotting).

Workaround I tried: converting the TAP results to QTable instead of Table (see astropy/astropy#17996 for details); using .unmasked for the arrays for plotting, trying to manually override the units, etc.