Robert Connell Clark and Mark Merlin also make a similar claim in Cannabis Evolution and Ethnobotany about high THC plants. They claim that plants will revert to low THC quickly within a few Generations if not continually selected for high THC.
I am a little bit skeptical of this claim, however there could be a basis if high THC is actually a negative trait on a population level.
This could make a lot of sense, since THC and terpenes are made from the same subunits of isoprene at some point in their biosynthesis pathway they are competing for reactants. Higher THC likely is correlated with lower terpene content. Since, terpenes may be protective against pests or pathogens, that would be a liability for survival and seed set. That assumes a lot about the function of cannabinoids and terpenes that we don't really know though.
So, to argue that a trait is unnatural because it is selected by humans really is especially moot in the context we are talking about right now. It is called Artificial Selection, and without it we would have wild hemp and no marijuana.