I myself found what the problem is:
I made changes to the model (for example like adding an RLC-branch to the model, or changed the topology of the model) after I saved the final values. So these saved final values which are going to be the initial values for the next simulation, cannot be applied, because I had changed the model after saving the final states.
So if somebody in the futire has this problem, maybe this experience sharing would be helpful for her/him