Well, in a certain sense, you COULD simply conflate "margin of success" as being equivalent to the "target number" in GURPS. In other words, some challenges are a "Simply roll under your skill, regardless of how much you succeed"; some might be "Roll at least 2 under your skill"; some might be "Roll at least 5 under your skill." That makes sense. This also directly scales with skills, since the higher a skill rating, the more likely you are to succeed by the required margin of success.
It's not always an amount
under your skill, though. Sometimes, for an easier task, you might be able to get away with "fail by a margin no greater than 2"; and it's been a few years since I've read the books, but I seem to recall that the easiest checks could allow you to succeed if you fail the check by a margin of up to 7. Effectively, you could gain a temporary bonus of +7 to your skill rating when you are performing an extremely easy task under optimal conditions, and this is the excuse for why a professional pilot might have a relevant skill rating of 12 or so - because that would shoot up to a 19 during routine operations.
Really, a roll-under system is great when your chance of success isn't going to be modified very often. If you're almost-always making a straight check, then it's trivial to compare the number you roll against the relevant skill rating and determine success or failure. If you're going to frequently adjust for difficulty, then an add-up system works better; for example, you could modify GURPS to run with 3d6 and add your modifiers in order to hit a Target Number of 18.
Comparison of numbers is easier than addition, and addition is easier than subtraction.
The place this most bugs the living daylights out of me for this is active defenses. It feels completely wrong that a defender doesn't have to account for an attacker's margin of success, they simply have make a standard success against their own fighting skill. So a guy with a 16 skill can roll an 6, and a guy with an 12 can roll a 12, and the clearly far less skilled defender still successfully parries. With two opponents in the upper limit of skill range (15+), winning a fight basically comes down to sheer luck. "Attack." "Parry." "Attack." "Parry." For round, after round, after round.
From a game design standpoint, there are good reasons why they do this. Now, the particular implementation fails for the very reason you suggest - it's far too easy to create a character with a high parry value - but that doesn't mean the idea lacks merit.
Basically, it comes down to putting a clamp on min-maxxers, and keeping defense relevant. If you're familiar with D&D 3.x or Pathfinder, then you know how attack bonuses eventually render AC to be meaningless - the Amulet of Natural Armor +2 becomes a joke item when it changes your AC from 22 to 24 and the enemy is swinging with +30 to hit. And that's a big chunk of the game which is invalidated. AC is really supposed to matter.
For the functional range of weapon skills in GURPS (roughly 9 to 18), the linked parry chance will always be significantly lower (a skill rating from 9-18 generates a parry value from 7-12). If you think that it's too common for a success on the attack roll to be negated by a simple success on the parry roll, then instituting a margin of success would end up with the inverse problem - success on the attack roll would
very rarely be stopped by any sort of defense roll. It's difficult enough to roll under a 9 on 3d6, but you'll hardly ever roll under a 9 by
more than the margin by which someone rolls under a 12 on 3d6.
Instituting a margin of success system would also greatly incentivize the sort of ridiculous skill ratings that tend to derail gameplay. With a simple success system, there's little incentive for raising your effective skill rating (after called shots) above 14, because each further point gives you a decreasing benefit - you won't succeed that much more often with a 15 than you would with a 14. With a margin of success, it's an arms race to boost your score as high as possible, because each point of skill buys you extra insurance against the defender having a good defense or rolling well, and you negate that chance on a 1-for-1 basis.
With a simple opposed check, your chance of hitting is equal to the chance that you succeed on your roll multiplied by the chance that they fail on their roll. As you increase your chance closer to 100%, the overall chance of success asymptotes out to whatever their chance of defending is. You still benefit from raising your accuracy, but there are diminishing returns, and the defender is never put into a position where they have zero chance of defending. Really, there just needs to be a cap on defenses to prevent ridiculous stalemate fights. (I would suggest something like 50% as the maximum chance of a successful defense roll.)