@rory-smith and I found a bug in the way the ROQ waveforms are being calculated when we add tidal deformabilites lambda_1 and lambda_2. We solved this issue by adding a couple of lines of code to the base_roq_waveform function. The error was caused because the waveform dictionary gets overwritten after this block of code: https://git.ligo.org/lscsoft/bilby/-/blob/master/bilby/gw/source.py#L517
To check that the updated version works, we ran two different tests: one with the current bilby version, and one with the updated version. You can see that the first plot (which uses the current version of bilby) results in biased constraints, whereas our updated version (second plot) solves this issue.
