Rate splitting multiple access (RSMA), as a flexible multiple access and interference management strategy, bridges the nonorthogonal multiple access (NOMA) and the space division multiple access (SDMA). In RSMA, users share a layer of common stream whose achievable rate is limited by the user with lowest channel quality. In this letter, a framework based on RSMA with user separation scheme is proposed, wherein the users are clustered into groups based on channel gain and the inter-group common streams are superimposed in power domain. Taking precoding and rate allocation into account, an optimization problem is formulated to maximize the sum rate, which is solved based on the successive convex approximation (SCA) method. To further enhance the performance, a user regrouping strategy is proposed by considering inter-group interferences, which is tackled with a simulated annealing (SA) method. Numerical results demonstrate that the proposed schemes flexibly allocate the rates to mitigate the effect of the above issue and achieve a significant sum rate gain in overloaded systems.