May we reach it? Or must we? In what time? With what probability?

Interactive Markov chains (IMCs) are behavioural models in which nondeterministic and stochastic choices are loosely coupled. Recently, we have presented a sound and efficient methodology that enables the generation of uniform IMCs from high level descriptions, such as statecharts. For uniform IMCs in turn, algorithms are at hand to quantify, for instance, the worst-case risk to run into a safety-critical system state within a given mission time. This paper discusses a disturbing gap in this methodology, which is originally rooted in differences w.r.t. state vs. transition labelling, and makes it unclear how safety-critical statechart states are being identified by the final analysis algorithm. We here show how the entire methodology can be lifted to a setting that incor- porates state as well as transition labels. This allows us to use state labels for identification of safety-critical situations all along the methodology, while transition labels are the principal means for a compositional construction. We discuss how both are affected during the methodology, and introduce a may/must labelling of states in order to recover timed reachability properties.