@kishna..........PyMOL doesnot work properly for complex small molecules as I have myself found many flaws in it......especially the align command is seq dependent so it ll not work for small molecules and super is not that much reliable........
Its not straight forward to automatically detect which atoms of a small molecule should be superposed. Pymol works well with the pair-fitting option where you select manually which pairs of atoms should be overlaid. I've only done this in the GUI though, I don't know if its scriptable.
I would recommend Ajay Jain's code 'SurflexSim' for this task. I have utilized it for the last 4 years to great benefit. Simultaneous exploration of multiple ligand conformers and reassembly via a fragment based method of 3D similarity. The pharmacophore hypotheses returned by this algorithm scores among the best benchmarks in ligand based screening. http://www.jainlab.org/