The model is built normally. However, I noticed that many residues in
the active site (which is localized in the interface between each pair
of monomers) that are supposed to exposed are turned inside out.
This is probably because your starting structure contains steric clashes
between the sidechains in the interface region, and the easiest way the
optimizer found to minimize these clashes was to bury the sidechains.
You could try building multiple models - each model will have a
different starting structure and so chances are that at least some of
them will not have this problem. Or you could take the .ini file
(starting structure) from your first run, and modify it manually
(perhaps with something like PyMol's sculpting) to put your interface
residues in a sensible position, to guide the optimizer towards the
desired solution. Or you could add some restraints to those sidechains
to keep them exposed - perhaps upper bound distance restraints between
Cbetas in the two chains.