I think they are referring to the fact that modulation is occurring at a lower frequency than the transmission.
You can modulate a lower frequency carrier and then increase or decrease the frequency prior to transmission with non-linear mixing with another frequency. This is commonly done to simplify the design. The modulator section operates at a fixed frequency and does not have to be retuned to change transmission channels. Only the mixer stage requires adjustment to change transmission channels.
As a simple example, lets say that I modulate a 10MHZ carrier and I want to transmit at 100MHZ. I could use a crystal oscillator to form a 90MHZ continuous wave signal and then combine the 10MHZ modulated carrier and the 90MHZ sine wave in a mixer. The output would be the individual frequencies, their sum and their difference. Since I am interested in 100MHZ, I then filter out everything but their sum. Now, if the next channel I want to transmit on is at 110MHZ, all I have to do is change the crystal and retune the output filter. The modulator remains unchanged.