Based on my tests I have found that the NFM mode (or FMN) on a Uniden BCT15 (the only one I own - not the "X" model but I would guess the RF and IF stages are either identical or nearly so) only engages audio compensation and has no effect on IF filter response (i.e. "adjacent channel selectivity"). In contrast, the GRE radios I own do exactly the opposite - no noticeable (by me) audio changes but a definite change in IF filter response. Ideally, it should be both but in both cases (Uniden and GRE) corners were cut for cost reasons I guess.
I do have considerable RF lab experience so I do know what I am talking about; however, I currently lack the proper test equipment to really definitively test the issue although I know how to do so if I did have said equipment. Instead, what I did was use my "trained" ears and experience and did the best I could.
Try the following as a test:
Tune to a strong continuous broadcasting and modulated signal such as the NOAA weather channel or a trunking control channel. Now listen to the audio quality and loudness with the NFM/FMN setting engaged versus not engaged. If the radio is doing some form of audio compensation for the deviation then it should be noticeable - wider deviated signals (such as the NOAA weather broadcasts) will seem louder and may distort in the narrow setting. If you cannot hear any difference then the receiver may not be doing any audio processing changes.
Now, to test if the receiver is doing any IF filter changes, try tuning off frequency the continuous broadcast by some amount (try small steps at first like +/- 5KHz or 7.5KHz and increase if necessary depending on how strong the signal is - experiment) so that you are still hearing the audio but somewhat distorted - ideally you want to be just at the last frequency edge of being able to detect the signal. Now listen to that distorted off frequency signal in both FMN/NFM mode and standard FM mode. In the NFM/FMN mode if a narrower IF filter is being used you should notice a dramatic change such as the signal completely disappearing while if no filter change is done you should either hear no change or some audio change (due to the FMN/NFM audio processing change).
Again, in my case, using only the BCT15 and comparing it to my GRE gear, I detect no filter change in the Uniden when using the FMN/NFM mode but some audio change while I get the exact opposite in the GRE gear.
I first really noticed this when trying to use the BCT15 for local state fire communications on the newer narrowbanded frequencies. It was a real problem because there were multiple fires around me and divisions from two or three separate counties were operating in close proximity (RF-wise) to me. Many of the newer channels were being used in addition to the older standard ones simultaneously. Essentially, the newer channels were useless to monitor on the BCT15 because the adjacent channels "bled over" so badly. I had to use the GRE radios which had either no problem or far less of a problem (maybe some slight "bleedover" but not so bad I couldn't "squelch it out"). As far as I could tell from monitoring the users communications it did not seem that they were experiencing any "bleedover" issues themselves (which you would expect using better professional gear). Unfortunately, this negated my intended usage of the BCT15 which is otherwise a good unit and has much better front end performance than the GRE's (in the lab, we considered the IF selectivity as separate from the RF "front end" performance so that is how I am looking at things here). I do not know for certain whether the other Uniden models are similar But I would guess they would be. I do recall a user (and admin on this site, I believe) who, in the California forums wound up thinking he was hearing communications on one of the newer "narrowbanded" CDF channels when in actuality it was determined that he was hearing signals actually transmitted on a channel 7.5KHz away and I think he was using a 996(X or T I do not recall) so I think that at least most of the Unidens use a common IF design.
I've said it before - I wish I could have a Uniden front end combined with a GRE IF section in one unit! Oh well...
As an aside, I performed the same tests as describe above on my Wouxun dual band HT which is a Part 90 FCC certified unit for professional use and found that it seems to do the same thing as the Uniden in terms of not really changing any IF filtering when the receiver engages NFM mode but only changing the audio (discriminator settings, I am guessing); however, in this case the standard IF filter was far more selective than the one in the Uniden scanner so that this was not really a problem. I suspect that cheaper professional models (like the Wouxun) might do this in many cases but I am guessing that more expensive higher end gear likely does have special selectable narrow/wide IF filters.
-Mike