A residuals plot produced by the pre-1.0 version of refl1d showed residuals all very far from zero in the same direction, how the χ² was unreasonably low for the displayed residuals. This dataset had an intensity parameter fit to 0.88.
It may be that the beam parameters are not being applied before the residuals are plotted in the mpl plots within the refl1d probe object.