Hi Simone,
Thank you very much for posting about this issue. The development team is aware about this for the sinestream signal and are working to fix it in a future release.
The other option would be to use the superposition or the PRBS signal within the block dialog in the tuning experiments. You would not run into this issue.
Thanks,
Sid
P.S. Additional clarification: the issue is that %conv signal is not computed correctly when te excitation signal is a sinestream. You can still use the Closed-Loop PID Autotuner block with sinestream signal, but do not use %conv signal for defining when to stop the experiment - define your own logic for that (for example, be conservative and just specify conservatively long hard-wired time duration).