In this paper we consider knowledge and algorithm engineering in combinatorial optimization for improved solving of complex TSPs with Time Windows. In addition to Nested MonteCarlo Search with Policy Adaption, as invented by Rosin (2011) and applied to TSP by Cazenave and Teytaud (2012), among other refinements to speed-up the exploration we perform beam search for an improved compromise of search breadth and depth and automated knowledge elicitation to seed the distribution for the exploration. We show promising results on TSPTW benchmarks and indicate improvements for real-world logistics scenarios by using a multiagent simulation system with each agent computing and trading their individual TSPTW solutions