Decomposition of a Semi-Markov Process Under a State Dependent Rule