Optimal route planning under threat of ambush is essential for any vehicle carrying valuable persons or objects. Our paper proposes a planning method for such a vehicle in an urban environment where, starting from a depot, a set of predetermined destinations must be served. We develop a previously existent flow-based single-destination model, proposing a method for minimizing attacker payoff in the multi-destination case. Our method requires specification of destination ordering, and we propose a method for mixing of these possible orders. Such mixing further reduces attacker payoff. We then apply the method to a set of possible scenarios based on Cambridge, Massachusetts, and analyze the benefits of our mixed ordering strategy. Finally, we introduce a second-level optimization model which again reduces the overall risk of successful ambush.