After all the latest model and fitting work by JeffK and Lilli, Jeff installed the appropriate inverse sensing filter (same as for 'usual' darm, but doesn't have the cavity pole in it, and also doesn't have some high freq stuff, since those get take care of in series just after the filter). 

I had been confused at that time, and wondered why the CFTD inverse sensing filter bank also had a whitening filter in it, so Jeff turned it off.  After further looking around today, I discovered that really it's just that the CFTD path does the whitening in a different place than in the 'usual' path for some numerical precision reasons, so we really did need that whitening filter on.  So, with the updated inverse sensing filter, and the whitening filter correctly on now, and also the integer number of delay cycles set to 9 to match the 'usual' DARM path, the CFTD version of DARM seems to be ready for use.

It looks like there is some discrepancy at high frequencies that I don't yet understand, but otherwise I think we can use the CFTD version for references when we are doing tests.