Jerome Pety sent me simulation results for the HCO+/YSO problem.
I added the results to my write up.

The main conclusion does seem to be that ALMA+SD+ACA performs better than
ALMA+SD alone. Also, the MEM-based result *so-far* appear better than the
CLEAN-based simulations, *but* the MEM images do not contain phase and
amplitude noise (Mark Holdaway is working on that now) so that *may* be
spurious (but we'll see soon enough).

I've also calculated % recovered mass, where again MEM seems to be doing a
lot better, and the improvement of ACA does not seem very large (e.g., 93%
of the mass recovered vs 89%), but as noted in the write-up, the slope of
the density profile is an important parameter when testing collapse models
-- much more so than mass -- and the deviation can get as large as 30%
(and no longer look like the nice monotonic fall-off that the input model
contained!). So expressing the results as % missing mass makes things look
better than they are (not good if we want to argue in favor of ACA).