Grafting using antibody.mpi.linuxgccrelease , the modelled structure does not have constant region of the antibody.

Generally speaking, Rosetta modeling tends to work with the Fv region of the protein -- for most modeling problems that's the region of the structure which is actually "interesting".

If you do happen to need a model of the full Fab or full antibody structure, my recommendation would be to look into combining the modeled Fv region with a known model of the rest of the antibody. Using something like RosettaCM, you can use multi-template "homology modeling" combine the two structural models. (The modeled Fv region is used as a template for just the variable region, and your constant region model is used for the template of the rest.) So long as there's a small amount of overlap (e.g. keep just the first bit of the Fv in your constant region model), RosettaCM should be able to graft the two models together.