Interval methods provide a means to implement strategies for simulation and parameter identification of dynamic system models with bounded uncertainty. The focus of this paper is the verified identification of continuous-time dynamic systems which are characterized by ordinary differential equations with non-smooth right-hand sides. Such models arise, for example, while modeling mechanical systems with a transition between different model states such as sliding and static friction. A novel interval subdivision procedure is presented for the verified identification of such system models. A comparison of numerical results for the identification of a drive train test rig concludes this contribution and highlights the advantages in comparison to previous work.