Is it possible to execute ab initio ligand docking onto a certain part of the protein (i.e. specifying a certain surface where to look for the binding pocket) without excision of that certain part which can result (or - in my case - resulted) in experimental artifacts?
It will depend on the knowledge you have for your ligand.
If speaking of a small molecule ligand, you would define it as active in HADDOCK jargon, and the define as passive the surface residues of your receptor in the region you want to sample.
If this would be protein-protein docking, you would need some idea of the binding site on one of the two systems.
And if using a local version of HADDOCK, for the random AIR scenario you can limit the sampling to specific regions.