The control subjects were abstinent for at least 60 and 90 days, respectively.
The article you cite mentions the following in the Materials and Methods, and I quote:
From the remaining subjects, age, gender, and AUDIT scores were used to create a matched control group reporting no marijuana use in the past 60 d.
A similar procedure was used for the 90-day control group. I agree with you that this method, i.e., at least 60-90 days of abstinence, leaves plenty of room open for variance. Probably a large part of the identified controls have not smoked marijuana ever, others maybe once or twice, but who knows? I think you have identified a very good point of critique.
The article mentions two citations (p.1506, Materials and Methods, first paragraph: Filbey 2008, Claus 2011) with respect to this protocol, and I followed these as I was curious if they actually controlled for this shaky part of their method to generate a control group. The first reference is actually not in the reference list(!) and after a thorough search I dug it up. I couldn't find any details on the questionnaire, however, and the same holds for the other article. If it is important to you you may want to read these articles in full (I just scanned the section on the questionnaire) to search for clues. And contacting the authors is never a bad thing either. For now, I conclude you are right and the methods are shaky.
However, flawed as mentioned in your question title is a big word. As long as you take the methods used into account, you can place the results in perspective against other studies that, perhaps, used study participants that have never touched any THC-containing product in their lives.