I think I can help about what the article means, thanks for posting it.
The simplest assumption here is that THC/CBD ratio is dependent on a single gene with two different forms, one which codes for THC, the other for CBD, and neither of which is dominant over the other. Some data is consistent with this idea, and some researchers have argued that this single gene model applies at least to an important degree. However, the problem is that if this model was accurate, we would have strains with only one of three ratios: 50/50 THC/CBD, 100%THC, or 100% CBD. What we see though is continuous variation in ratios among strains, which can't result from a simple single gene control of the ratio. It is clear that while a single gene may be important, its influence is moderated by other genetic mechanisms. As usual, the simple idea suffers in the face of reality.
I think that the bottom line is that we don't entirely know how cannabis does it, but it does, and breeders can exploit this whether or not we know how it works.
Just my two cents worth based on my limited reading of this stuff, I am sure that there is a pile of stuff out there that I have not seen.
I can add some more to this since I looked into it a while ago, there are some more specifics known already and described in scientific literature.
de Meijer indeed proposed that model with the Bt and Bd allele of a single gene.
this was supported by crossing experiments, this showed the expected ratios:
thc x cbd= all 50/50 ratio f1, and in the f2 25% thc, 50% 50/50 and 25% cbd.
newer research went more into the actual sequence to find the thc and cbd genes. it was found they are not actually two alleles of the same gene, but two different genes (different locations). this would make it theoretically possible to get a 50/50 ratio plant that stably inherits that chemotype (no re-appearance of full cbd and full thc plants). however, these 2 genes are very tightly linked. physically they are somewhat close but there is some distance, but they are in a region with minimal recombination.
so practically, that model from de Meijer does mostly apply.
now the point you bring up about continious variation is where it gets really interesting. if you look at the thc/cbd ratio, not absolute amount, you can clearly distinguish those 3 classses (full thc, full cbd, 50/50). but, there is indeed quiet some spread within those categories, even if they do form 3 clear groups.
which is quiet relevant for legal cbd production, since a full cbd plant will still make some thc. so if you increase the total cannabinoid content/resin production, the ratio thc/cbd will be the same, but the legal limit is an absolute number. so let's say you start with a 8% cbd, 0.2% thc strain. <0.3%, legal to grow in america. you breed for increased resin and manage to get it up to 16% cbd. but now your thc% has doubled too to an illegal 0.4%, since the ratio stays the same when the absolute cannabinoid content increases.
so where does it come from?
first some explanation of how thc and cbd get made.
like a typical biosynthesis pathway, there are all kinds of intermediaries, and enzymes at different steps to do stuff and get to the next step.
one of the last steps is the production of cbg (well, actually cbga, just like fresh plants contain thca instead of thc, but I'll just leave that out).
cbg is then the substrate from which cbd, thc and cbc get made. (and maybe more minor cannabinoids).
so going back to those 2 alleles or 2 genes, that would then be different enzymes, one which can make thc from cbg, the other can make cbd from cbg. but apparently no thc-enzyme is needed to produce low levels of thc. this could have multiple explanations:
-there could be multiple copies/pseudogenes of the thc gene, some of which might be totally broken, but some might just have their expression turned really low (or the enzyme is really inefficient but still can do something), resulting in a low production from those copies even while the main thc gene is not present. (or, the main thc gene could still be there bt in a deactivated form, which could similarly be not fully decactivated. but I think, if I remember right, that it was found that hemp plants don't contain a thc-gene at all, while drug-type plants do contain a broken cbd-gene)
-it's in the main enzymes themselves. this option I think is especially interesting, since there is some evidence for it. the cannabinoid synthases are said to be 'leaky', i.e. the cbd synthase makes not just cbd, but also minor amounts of other cannabinoids, including thc. the cbc-enzyme gene was actually first described as an inactive thc gene (so, that would be such a copy from above), and later it was found it did actually do something, but it made mostly cbc instead of thc. people also found that if they put those cannabinoid enzymes into yeast, they could produce different cannabinoids depending on environment (don't exactly remember which factors they changed, I think growing the yeast at a different pH or temperature).
