There appears to be too much useless "noise" at the discriminator level on the circuit board that should be filtered out with a ceramic capacitor of fitting capacity.
I discovered during the process of installing a discriminator tap in my Uniden BCD996XT that including some tap line filtering really helped with my computer's ability to decode digital trunking system signals.
I limited the max. audio frequency on the tap output to around 20kHz with a combination of a tap line resistor and a ceramic capacitor. In my experiments this markedly improved my computer software's ability to decode the P25 signals on the line. I have the impression that I could probably filter down more aggressively to cut out unneeded high audio frequencies that are not part of the signal of interest.
Of course, the Uniden BCD996XT decodes P25 phase 1 without the need of a computer, I just ran the above setup as a research project to see what effect cleaning up the discriminator level audio has on digital signal decoding. And it really helped. If I could figure out what the internal discriminator audio line impedance is on the Uniden scanner board then I could just do a quick filter calculation and pick the right capacitor to solder onto the discriminator test point against ground.
I have not had time to do more research and play with my gear, but I am starting to think part of what I have been having issues with is front end overload/adjacent channel from a number of Land Mobile/Cell sites along with a AGC system that nobody appears to understand/use as it is questionable if it works.
I really think if the AGC worked and could be properly configured and enabled that this might help matters along with something the poster above mentioned about how the filtering is for the P25 decoder.
I also speculate that a more stable base clock might help as well, but I have been totally brow beaten, discounted and bullied for even mentioned that there could be anything that could or might improve matters short of a full receiver/decoder design from the ground up.
But all this discussion is useful, out of the box thinking and possible trial and error may actually yield an improvement/hack/mod that might make decoding just a slight bit better.
Kind of like the 536HP headphone issue, maybe it will turn out the wrong capacitor or resistor tube was installed in the component placement machine and swapping a few parts around could improve things?
You never know until you try.