Not enough precision in results

[phannebohm]

The reference results probably need to be generated with stricter tolerance. See OpenModelica/OpenModelica#14124

[casella]

We should also remove the derivatives. They are polluted by numerical noise and normally not very interesting.