There seem to be a few issues related to the normalization of the Pl matrix.
The normalization are not applied for TB and TE spectra (the
P02 term) in here and here. Assuming the normalization is the same as other terms of the matrix, this should fix the bias in the TE/TB spectra.
_bin_operators function computes the binned
dSdC is for
Cl, this causes a biased signal matrix. Disabling
_bin_operators makes the estimated S matrix match with MC simulations. But this only works for single multiple, I'm not sure if the fix should be the other way (correct the dSdC -> dSdD) for binned spectra.
To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information