Same thing here in the western suburbs of Philly on 1620 on the expanded band. 1060 KYW + 560 WFIL = severe intermod on 1620. The transmitter sites are about 2 miles apart. Surprisingly, we have never found such issues on 1550 (560 WFIL + 990 WNTP); I think that's because of the heavy filtering of the two stations on our diplexed transmitter site. The reason I say this is that the 1520 - 550 intermod surprises me. I would tend to doubt that, unless there are issues with their diplexing/filtering (which I doubt), the problem in Buffalo is much more likely due to receiver generator imaging.
I have seen some wierd things happen over the years, however. In New York, when I was the senior radio engineer at WBBR 1130, we got a frantic call from a ham in North Carolina that he was hearing us clearly on the 160 meter ham band. Our transmitter supervisor, Bob Janney (who is one of the best transmitter guys in the business) and I immediately went out and did analysis of our signal armed with field strength meters, a spectrum analyzer, and several portable radios for comparison. Sure enough, loud and clear on 1830, there we were, but only at night when we were on directional. It took a couple of days of monitoring and signal chasing but finally we found the problem -- but it was not with our transmitter or antenna system! The field strength meters showed that the signal was not coming from our site, but from WZRC 1480's transmitter, located about 3 miles north of WBBR's. When they would switch to their night pattern, they would begin transmiting a pretty good copy of WBBR's audio on 1830. 1480-1130=350 khz., Take that difference and add it to the 1480 carrrier and - voila - 1830! We worked together with WZRC's engineer and finally found a bad capacitor in WZRC's night phasor that was generating a 350 khZ signal with WBBR's audio and feeding it back into their transmitter to produce the 1830 signal.
So, hey you never know what can happen when you're around a lot of RF>