Or we can have some humility and admit that NHL amateur scouts, most of whom have years if not decades of high level scouting and playing experience, watch each prospect play in person dozens of times, have the opportunity to meet with and interview the player and his teammates, coaches and family, have access to medical reports, etc. are better in the aggregate at evaluating prospects than a model that seems to rely exclusively on points.
If we start from there then it's obvious that someone who was passed on 122 times by NHL scouting staffs in a draft that happened less than a year ago probably doesn't have "a good chance to succeed." This is of course supported by the data on actual pick success rates which is not a model but an objective historical record. We can also look at the current landscape of NHL defensemen and notice not a single one is as short as Cagnoni.
None of this is to say there's a 0.0% chance Cagnoni has a NHL career but 65% is so laughable as to completely discredit everything that guy puts out in my eyes. The ridiculous Merkley projection is just another of many examples.
There are no perfect methods. Statistical models are wrong in part because of the reasons you listed, along with their limited scope but they do not only rely on points. Scouts are wrong because they overvalue one trait and undervalue another, or get a bad vibe from a player due to their own subjective analysis (See Tim Burke and Anze Kopitar.)
The infinite number of variables that can make or break the development of a career are not not quantifiable, by a scout or a statistical model.
Appealing to one flawed method as a means to discredit another flawed method is not a valid form of proof. It is also why you can't dismiss either method. They are a means to create the best picture that you can of a player given their development to that point.
Merkley is a perfect example of this, where the statistical models loved his play, and only his play, but his personality and other various issues absolutely derailed his career. The statistical model is simply saying that players with similar statistical projections made the NHL a certain percentage of the time. If you interpret that prediction as the full evaluation on a player then you as the consumer are dismissing the other factors involved, and that is not a strike against the model.
There's no need to be dismissive Byron's work or the work that goes into those statistical models. Those people work hard to create them, and while they're built on an imperfect science, because the games are not played on a spreadsheet, there is value in the patterns that emerge.
Also, the irony of you appealing to the authority of NHL amateur scouts when you have a severe personal vendetta against one, is not lost.