As a game designer, imagining myself to be a player for the sake of understanding how the game will play out, I absolutely don't want the mechanics to dictate that a character is bad at whatever thing they're supposed to be good at. There are plenty of games where you try to build a competent marksman, and the mechanics dictate that you'll still fail to hit a barn door more than half the time. Those are bad games. I didn't sign up to play Keystone Cops. Early D&D was notorious for this, with the Thief class that had a pitiful chance to do anything.
If combat in Pathfinder was nothing but one character on the receiving end of 20 arrows, then you might have a point. As it stands, the tank being targeted is a fairly small part of the over-all session. When it does come up, I expect the tank to succeed at their task, in much the same way that I expect the rogue to successfully disable any traps we find. That's the entire reason why we brought them along in the first place.
Armor Class matters for non-specialists. The ranger has a better AC than the wizard, and while neither of them should be on the receiving end of too many attacks, that's exactly the kind of unpredictable situation where we'd expect the variance of the die to matter. The ranger might be able to pick a lock, of the rogue is indisposed, and that's why we bother tracking all of these numbers. But one a specialist is operating in their area of specialization, they should succeed an overwhelming majority of the time.
Which D&D are you talking about?
Early D&D (or OD&D) if you wanted a character to sneak past a guard it was DM's choice. In many games that meant a Dexterity check.
If the character was a thief, they got that dexterity check ON TOP of their other checks.
Just because one wasn't a thief did not mean that they could not actually use their ability scores to try to sneak past, steal, climb up a cliff with a rope, or multiple other items.
It was a later iteration that caused players to think this way (as it did not really notate this in 1e so some really weird people had a rule that a fighter could not walk quietly and could not climb things and other such crazy notions).
This got further reinforced with 2e...
But luckily nothing in the rules PREVENTED ability score checks (and prior to Non-weapon proficiencies ability score checks were actually encouraged, similar to how 5e handles many of it's skill systems, but less structured).
However, 3e I think sort of made this an even worse exaggeration of skills and such and it only started to change with 4e (which handled things similar to 5e but all around with a +5 to skills in general (instead of the +2 to +6 proficiency spread).
Your idea doesn't really hold water with the early thief class and how it was handled...though it probably holds water with LATER AD&D 2e and especially 3e.
As far as making the Thief a marksman...the crazy thing that people expect now is that some untrained lackey is going to have the same ability as a trained warrior. That a soldier is going to be just as proficient at hitting a mark as a guy that spends his days buried in a book, or a burglar who spends his time sneaking around.
Early D&D didn't have this ridiculous illusion of everyone is equally good with weapons and that a marksman who actually was a MARKSMAN (aka...a fighter that is trained in weapons and combat rather than some guy who uses weapons in his spare time adventuring but otherwise is a priest or a bookworm) and would hit better than others.
At level 10 he probably could hit an AC10 target pretty consistently. Even one who was at 5-7th level could probably hit an AC10 target pretty consistently, and probably hit their mark more often than others.
