A Hybrid Approach to Reasoning with Partially Elicited Preference Models