Hi Pico,
I have found the same results! And I have a theory about why.
For others: I found that dmpGetGyro gives rates of Roll Pitch Yaw (note the order) in counts that correspond to +-2000/25 = +-80deg/s (divide count by 32768/80)
My configuration: DMP FIFO set to 25Hz (see note below). getFullScaleGyroRange says 3 (2000deg/s), and I noticed it is set to this by dmpInitialize(). getFullScaleAccelRange says 0 (+-2g), but as others have pointed out, this does not seem to reflect the DMP output (dmpGetAccell), which is requires divide count by 32768/4=8192 implying +-4g. (getAcceleration() has a note suggesting 8192 is 2g sensitivity but I don't understand this.)
My first impression of the angular rates scaling is that it includes the fifo frequency I am using of 25Hz. Luisrodenas mentioned "multiplied by its gain" but I don't think this is what he meant. I have yet to explore this by changing the frequency, but I am now skeptical because you (Pico) are likely not using 25Hz.
For anyone finding this wishing to change your FIFO rate, do a search in the MPU6050 header for inv_set_fifo_rate and read the note below it.