There are foreground residuals that are significant relative to 5e-4, see pages 15 and 18 in the pdf.
There is strong sensitivity to the higher frequency bands, see page 32.
SH: how is S4 doing with this model
MR: Need to check the maps and the mask being used. Perhaps can optimize the mask, but smaller patches will lead to higher noise. Worth trying.
Is the model even consistent with current constraints on the sky?
Question: slide 15, why is there a break in the noise at \ell= 700
Note: MR had done r=0.003 only on models 91,92 so far