How should thermodynamics for small systems be done?

We show that the question of which simulation method is the right one to describe the thermal behaviour of clusters is not a philosophical one. Moreover we point out that getting better results might be accompanied by a loss of information. Different points of view yield different results. The number of iterations, let us call it simulation time, plays a central role here. Making use of the simulation data at higher temperatures the phase transitions and the effects of isomerisation are easy to understand.