Investigating Intra-Abstraction Policies For Non-exact Abstraction Algorithms