A note on computing the tail decay of M/G/1-type Markov renewal processes