Flex has been doing some upgrading on the DSP in recent software releases, so I did a quick video demonstrating APF in the 1.4.12 version of the software. It's important to list version since the software is continuously upgrading and what may be true this month may be better next month depending on what's been improved.
The present APF has a slider which gives between 3 and 12 dB control of gain when APF is selected, but the APF precedes the AGC loop. If you want more gain beyond APF you simply increase the volume control.
In my tuning I usually set the AGC-T to a "best" level for a given set of band conditions. AGC-T sets the level where the AGC starts to react and reduce gain. For weak signals the sweet spot is to not have the AGC-T too high where AGC is fully engaged or too low where the signal is below receive threshold. This is useful because it tunes the radio to a given band/antenna for best performance. For strong signals you just let the AGC reduce the gain. Once set for a band there is not much need to "ride the gain" as I have seen others complain. If you are talking to some really loud stations you can turn the AGC-T way down and band noise is virtually completely eliminated.
The effect of APF is more evident when used with wider bandwidths. The control was introduced years ago as a means to get better bandwidth in the days of crystal filter radios. If you went from 500hz to 250hz to 125hz crystal filters bandwidth reduced but distortion due to ringing and delay went up dramatically. In a SDR radio where the filters don't ring APF is less useful, but still can provide a way to better tease intelligence from the noise, and this is the bottom line. What we want is improved signal to noise.
In the video I set AGC-T where I like it and AGC to fast. I next set the band width to 250hz and tune in a typical 15M CW signal. We are listening to AO150A the ITU special events station knocking off all comers. The band noise at 250 hz is averaging -112 dBm and at 50hz is -120 dBm. AO150A is about -110 to -105 dBm according to QSB. I try various demo's of APF trying to switch APF off and on during his transmission. Midway in the video I switch from fast AGC to AGC off. This sets the stage gain at a fixed level and makes the APF act more like how APF would work if it followed the AGC loop. I run through similar bandwidths and off/on scenarios.
I think if you focus on the S/N and noise reduction provided by APF it is quite effective especially in wider bandwidths. As you reduce bandwidth a great amount of noise is already eliminated and in that scenario APF is less effective.