If you have indeed a buried binding site you can run with scaling down interactions for rigid-body docking
Guru interface -> Energy and interaction parameters
(or for a local version, search for kinter in run.cns)
Scaling of intermolecular interactions for rigid body EM -> 0.01
And then change also the weight for the van der Waals it0 scoring (w_vdw_0) to 0.
I would also suggest to run with standard parameters (kinter=1.0), but upscale w_vdw_0 to 1.0
And take the best results of the two runs.
From our experience, it might be better to look at it1 models rather than water for small ligand (especially if hydrophobic and buried)
Also set the number of MD steps for the first two stage in it1 to 0.