Thank to Steve Knell,
I've found a bug that arises in gqrx wrapper when the tuning is done clicking on the waterfall.
The problem is related to the fact that gqrx uses a channel differential frequency to not move the hardware frequency.
This is perfect, not gqrx problem, but from TCP/IP socket only "real" frequency is avalable, and the wrapper uses this frequency to set the Center frequency on the device (DUO; S1, S2).
The tuning works well if youj don't use the "clik on the fft" to set frequency, in other words when the differential channel frequency is zero.
I've cheduled to change the communication between wrapper and gqrx using I/Q channel two way (as gqrx do with devices), but it takes time and effort, so I ask to be patient: even if it is not visible I'm working on this.
Regards,
Giovanni - HB9EIK