To make this model we fussed around a bit with strand pairing to try
to get the alignment to produce a sheet with consistent exposure of
the charged residues.  We also had to tuck the initial and final
helices out of the dimerization interface and what we believed to be
the DNA-binding face.  

We played around mainly with a monomeric model, since undertaker only
understands monomers, but this did not seem to cause much trouble for
us, other than for the initial and final helices.

When we'd fussed with the model to the point of diminishing returns,
we made a dimer (aligning to 1h0x) and repacked the side chains
(without moving the backbone) using Rosetta.  We think that this fixed
a couple of bad rotamers at the dimerization interface.  The K78-T81
hbonds look particularly good in the repacked dimer.