Thanks again ... your suggestion works in principle, however, it does so only if the three template structures, including the mutual orientations/locations of the three domains (two Fab, one Fc), represent a physically plausible structure. If I use as templates each of the the three domains with a correct conformation, but a random orientation/location then modeller leaves the domains close to their initial location and generates a hinge with some of the bonds way too long - same outcome as when I use IMP/rrt_sample with a comparable input.
I was trying to find the source of rrt_sample, to see whether this can be fixed with some minor modification of the code, but I failed to find it - rrt_sample appears to be included only as binary - would you be ready to share the code?
Best regards,
Michael