The filter responses to be assumed here are Gaia G and Bp, and their responses truncated at 380nm, where the PFS coverage ends. See the plot here.

Using the Pickles stellar library and limiting the stellar type to A-G stars, I got this figure. As expected, the Gaia G-band response at <380nm only has a small effect and the magnitude difference in the truncated G-band filter is only 0.017mag. The scatter is small; 0.003mag. In practice, the scatter will be even smaller because we limit the FLUXSTD spectral types to a narrower range (i.e., we use late-A to early-G only). So, I think we can safely use G = G_truncated + 0.017 for FLUXSTDs.

As for Bp, the effect is larger; the median magnitude difference is 0.1mag. However, the scatter is still only ~1%. I could fit polynomials to the points here, but I am not sure if that is needed. Maybe the simplest solution is just to do Bp = Bp_truncated + 0.101.

OK, I fitted polynomials just in case they are useful.