I understand what your saying. Its true that environmental conditions can lead to a loss of overall genetic variability, only the strong survive, but those same conditions weed out weak genes, ensuring that only the offspring of the strongest survive. That's how natural selection works. If you had an island full of white tigers, then one would be correct in stating that the dominant color gene for the population of tigers on the island, is white. The orange gene may be dominant genetically, but if a dominant gene does not exist in a population, then it is not dominant. The gene for having six fingers is dominant in humans, that is a well known fact, how many fingers do you have?
Lets say you dropped a population of cannabis plants, (containing both photos and autos) into an extreme environment, say Alaska, and just let them be. If the seeds sprouted at the right time of the year and it was an early blooming enough strain to finish in time, then the photo plants might have a chance, but the autos would definitely finish in time to drop seed, more importantly, pollen would be swapped and open pollination hybrids would be created. Now consider the F1 generation of these hybrids, 75% of them would not autoflower, but in the case of a cold spring or fall, they would also not drop seeds and thus they would not contribute to the next generations genetics, but the autos would thrive, completing multiple lifecycles in a single season. Photo plants would occasionally pop up still because the gene is dominant after all, but much like the six fingered men that ought to be running wild somewhere, they would be freaks, and due to unfortunate environmental factors, they would have a very hard time passing on their freakishness. One can see that it would only take a few years of adverse condition to completely remove the dominant photo gene from a given population, it would not make the photo gene recessive, but it would be fair to say that the population in question is dominantly auto.