In this work we consider route planning for valuable delivery in an urban environment under the threat of ambushes in which a vehicle, starting from a depot, has to serve a set of predetermined destinations during the day. We provide a method to plan for hardly predictable multi-destination routing extending a minmax flow-based model available for single-destination cases. We then formulate the process of selecting a visiting order as a game to obtain a mixed routing strategy. We analyse the application of the method to a set of simulation scenarios and compare the mixed routing strategy against the best routing. Finally we develop further the methodology introducing a second-level optimization model which reduces the overall risk associated to the proposed mixed routing strategy.