Thank you for your suggestion, but the problem is, that the calculated losses would be much to high.
I found one value for the turn-off losses at 25 °C in the data sheet under characteristic values at the nominal point. At this point I devided the loss values at 125 °C by those at 25 °C. Then I devided all values at 125 °C by the resulting ratio and took the results as the loss values at 25 °C.
Now the simulation works fine.