You've hit on my number one concern when it comes to the current Uniden scanners - narrowband FM IF (adjacent channel) selectivity (NOTE - I am NOT talking about RF front end band "selectivity" here - I admit that Uniden does have a well known advantage in terms of this aspect of RF performance compared to the GRE units). I obtained a BCT15 and initially intended it for CDF (CalFire) monitoring; unfortunately, I too found out that the FMN mode did not kick in a suitably narrow IF filter and, just like your (the OP's) experience, I heard adjacent 7.5KHz channels all too well. I now have relegated that otherwise great analog scanner to listening primarily to CHP lowband stuff that is spaced at 20KHz increments so it works ok on those.
What is interesting to note is that the current GRE digital capable units (PSR500 and PSR600) as well as their Radio Shack equivalents (PRO-106 and PRO-197) do seem to engage a narrow IF filter when NFM mode is selected. Unfortunately, in these models, there are severe limitations for the use of this narrow IF filter. Firstly, you can only engage that filter for conventional analog frequencies programmed into memory only - you cannot engage it when in search mode, tune mode, conventional digital mode, and any trunking mode (these last two modes do allow you to set the mode to NFM via software or directly but I have verified in my own testing that they do not actually do anything so are essentially ineffective). Secondly, my own usage and testing confirms that there is a slight delay to the switching in of the narrower IF filter during memory scanning when these channels are scanned. If you have a channel programmed with NFM in a memory and that frequency has a strong adjacent channel interferer then you will notice that the scanner will stop on that channel and you may hear a slight "pop" sound as the narrow filter engages after which the scanner will hear nothing and will stay on the frequency until the delay time is up and move on - this can be annoying at times. Judging by these issues, I rather think that the GRE narrow filter was likely a late addition to the design - almost an afterthought. Still, at least these units DO have a narrow IF filter to use, albeit with some odd issues.
The problem seems to be that Uniden and GRE took two different approaches to the NFM 2.5KHz deviation mode accommodation. Uniden seemed to attempt to accommodate the lower deviation mapping to demodulated audio loudness (think Carson's Rule and modulation index) with their NFM mode but decided to forgo any attempt to design in a companion narrower IF filter. GRE, on the other hand, took the opposite approach and made their NFM mode utilize a narrower IF filter while no attempt was made to automatically engage any sort of 2.5KHz deviation audio amplitude mapping (you can engage a "6dB audio boost" manually but, in my mind, that is a severe cop out band aid solution).
Ideally, both manufacturers SHOULD have designed in BOTH of these approaches and addressed the new narrower narrowband FM 2.5KHz mode as it should correctly be from an engineering design standpoint. It is truly odd that they each addressed HALF of the problem and each addressed different halves. I simply cannot imagine that the cost savings of accommodating only half of the 2.5KHz FM deviation issue is worthwhile and significant. I do have a background in RF design and do know what I am talking about - this is one of my major pet peeves in terms of Uniden and GRE's design compromises.
I was waiting to hear if the newer XT models of Unidens had added a truly narrow NFM filter to their designs and, based on the OP's experience, it appears that they have not. That is unfortunate. I agree with the OP - when more narrow 2.5KHz signals predominate and start appearing on adjacent channels these issues will start to become problematic. I think that this will mostly be noticed on the US civilian VHF high band as in this band the channel spacing will be 7.5KHz. In the federal portion of the band and in the UHF and higher bands the spacing will be 12.5KHz so it will be less of a problem.
-Mike