When we tried optimizing our initial model to improve burial and
packing, the terminal helices kept flopping over onto the dimerization
interface.   We had to add constraints to pin them to the back of the
monomer, since undertaker does not have the ability to optimize dimers
yet. 

This model (try15-opt2) is just a bit "foamy" and the helices are not
tightly packed against the rest of the portein, but we've reached the
point of diminishing returns and need to spend our effort on other targets.