Backtracking a pharmacophore hypothesis

Hi,
Given a 3D structure of a protein-ligand complex, Since it should be possible to generate a hypothesis based on protein structure by the "Interaction Generation" protocol and then, using the ligand, enumerate a hypothesis by "Feature Mapping" protocol, is it conceptually correct to validate a new hypothesis by combining these two (Protein based and ligand based)hypothesis and enumerating important features thus?

Also, trying to recover a native ligand extracted from the protein-ligand complex seems to be a little elusive. My work example is 2pr3. After generating a hypothesis through 'interaction generation' I tried to recover the native ligand (2371001)from an sd file.

Did i miss something here?