Most of microsimulation tools used to model roundabouts encompass classical gap-acceptance algorithms to represent the insertion of approaching vehicles into the circulatory roadway. However, these algorithms fail to reproduce the mean priority sharing process experimentally observed when the circulatory roadway is congested. This paper fills this shortage by proposing an integrated microscopic framework with: (1) a gap-acceptance algorithm giving relevant capacity estimates in uncongested regime; and (2) a probabilistic rate-based insertion decision module in congested regime. In this framework the car-following model can be implemented independently of the insertion decision-making process. Moreover, its direct influence on the insertion decision model is released in congested regime thanks to a relaxation procedure. The obtained simulation results are convincing compared to on-field data collected at different sites for both peak and off-peak periods.