Many techno-economic studies for the pre-design of local energy systems rely on mathematical programming. This comes with several modelling choices including simplified technological and economical models, temporal and spatial resolutions, or perfect foresight assumptions. On the basis of a single case study with complex decision making due to interdependent time scales, this paper shows the importance of evaluating and comparing the (cross-)impacts of the different modelling choices. The modelling choices tested are the inclusion of flexibility costs and constraints, the use of representative periods, the use of different methods to optimise operational decisions (including various rolling horizon methods), and the consideration of forecasts errors or not. The results illustrate the impact that flexibility costs and constraints can have and how they determine alternative modelling choices. They show under which condition representative periods can be used without introducing strong biases on the results. Finally, different rolling horizon strategies are compared, concluding on the validity of the perfect foresight assumption on the case. Results are discussed in terms of performance, validity, and computation times.