There are two sets of maps in this directory: 150 GHz (SZ signal + CMB) and 220 GHz (almost no SZ signal, CMB).
Both are in units of uK_CMB, such that the CMB signal is the same amplitude in each.  The weight map is a map
of the instrument's relative depth per pixel (it's nearly uniform in the central region), and the xfer map
is the transfer function (equivalent to the PSF) that takes into account both the beam and the time-ordered
data filtering.  

For just getting a quick look at the data, there's also a subtracted, smoothed map which was produced as
described in the paper--by applying a matched filter to extract the CMB signal from the 220 GHz map,
subtracting it from the 150 GHz map to produce a minimum-variance map of the SZ decrement, and smoothing 
with a Gaussian to remove the high spatial frequency noise.  The positive ampiltude "wings" to the east
and west of the cluster SZ decrement are an artifact of the time domain filtering: a low-order polynomial
is subtracted from the time-ordered data of each detector on each azimuthal scan, which is necessary
to remove the gross atmospheric variation.  

I can also give you noise estimates in the Fourier domain if you wish to attempt to fit models to these
maps.

Cheers,
Tom