Let me see if I can clear this up.

In a superheterodyne receiver the signal from the antenna is slightly filtered then mixed with a signal that is offset from the signal of interest. If the receiver is single conversion, the frequency is offset by 455 kHz as set by the crystal. This frequency can either be above or below the transmitter signal, it doesn't matter, as the mixing stage produces both the sum and the difference of the two frequencies as well as many other combinations of these two frequencies in harmonics. The reason for 455 kHz was that the IF parts ( tuned intermediate frequencies) were cheap common parts used to build AM radios. Unfortunately, the closeness of other strong signals getting through the lightly filtered front end at the antenna meant that you often go shot down in areas with a lot of other signals.

So the next improvement in RC RF was the addition of a second mixing stage that was offset by 10.7 MHz, which happen to be used in cheap FM radios. So what this really did was make the slight filter at the antenna work quite well at rejection of image signals, that it could not do very well when they were only 910 kHz away. Signals that are 21.4 MHz away are attenuated fairly well. Again, it doesn't matter if the frequency of this oscillator is above or below.

Now radios produced after Frank Hover developed the narrow off spike that made AM RC sets work without the AGC circuit going nuts allowing any stray signal to mess up the pulse train, RC development was off and running from the mid-60's and all was good until the mid to late 70's with AM radio sets. But thanks to cheap FM chipsets, it got a lot cheaper and easier to build FM RC sets. They didn't work any better or worse, but just cost a lot less to build. However, at this point the manufacturers had to decide if the narrow off pulse was a high side frequency shift or a low side frequency shift.