This is whole-try11-chimera, our best whole monomer, which came from
optimizing 81-end, trimerizing it, trimerizing a crude model of the
collagen-like domain, manually piecing the two trimers together, then
extracting just one chain.

At that point there were still some pretty serious problems with the
globular 81-end part, probably coming from a sheet constraint that was
misaligned. A chimeric model was made of the globular part, taking the
best parts from that model and from the original fully-automatic model.
The globular part was reoptimized and stuck back into the whole-chain monomer.

The angle of the collagen-like part was repositioned, but we did not
have time to get it quite right---as a result the monomer does not
form trimers simultaneously in both domains.  This could be fixed
fairly easily, but we were running out of time.