Wilks is really only used to determine the best male and best female lifter across all weight divisions on a given day. When you apply it to totals of competitors at the elite, ie international, level, it works out to be valid, because those lifters are much more likely to be lifting at a weight which suits their height and levers. The actual regression model was based on historical world champ data going back many years. If you look at the trends from world records, for example, you'll see that supers aren't actually that much stronger than 110s.
The formula was based off totals and was really only ever intended to be used for totals. It happens to work ok for raw bench only (although has been thrown out a bit since the advancement of shirts from the late 90s), but was not specifically designed for this purpose. But it definitely does not work for squat only or deadlift only, nor does it claim to, because the formula was derived from totals.
It is also not suited for comparing an individual lifter's efforts over his or her career. Wilks does not even come into play unless you are good enough to win an overall comp, so aspiring lifters are better off just trying to improve their total and getting towards their ideal body weight (by which I mean ideal powerlifting bodyweight, not shitty medical BMI recommendations!)
