During the last few months, considerable effort has been spent on assessing the post-CHF wall heat transfer package of RELAP5/MOD3/v7j (R5M3) as well as on investigating the effect of some model differences between this code and its predecessor, RELAP5/MOD2.S (R5M2). In this work, we shall outline the problems associated with the post-CHF wall heat transfer models and logic of R5M3 (which are partly responsible for the totally unphysical code predictions during reflooding), the main differences between R5M3 and R5M3 and we shall show that by implementing in both codes a physically realistic and sound wall-to-liquid heat transfer model, one can predict very well experimental results obtained in separate-effect bottom flooding tests.