The first model includes the potential cluster variable conditioned on clustering variables already selected, while in the second model the variable is not included.