But doesn't the above describe exactly how melee (or ranged) combat works?
It's close. But there's differences:
#1: You're expected to hit with an attack roll (or a magic attack roll). Your bonus to hit is almost always going to outweigh an enemy's defenses.
#2: In 4e, now we've got things that trigger on a miss specifically because whiffing is dull. In earlier editions, spells had effects even on a successful save. Or you have the phenomena of 1st level mages with auto-hitting magic missiles outclassing fighters with greatswords.
#3: Because you are expected to hit, the powers are not vastly more powerful than any other attack at the same level (if it's "balanced," anyway). In a "low magic" kind of setting, magic with a failure chance tends to be capable of things that aren't possible without it, ramping up the effects power (though, of course, not necessarily).
#4: Psychology matters. It's a much different feeling if my enemy dodges my blow than if I just fail to attack in the first place. It's kind of a semantic difference, because the effect is the same, but a "failure chance" is a "you suck at life" mechanic, whereas rolling against a defense is a "your enemy is powerful!" mechanic.
#5: You can make many individual choices that affect the chance for an attack or a spell to hit. You control that chance, to a degree, and that control is dynamic, it your chance can change from moment to moment. This means that while you might fail, you can always manipulate odds to be in your favor.
I would posit that it's a cornerstone of D&D and nearly every other game system that I can think of that you are entitled to attempt something each round but you're not guaranteed any results. My view is that the chance of failure makes success all the sweeter.
But you are
expected to get results. Sometimes you might not, and that's cause for cursing and dice-throwing, but most of the time, you will (even if you don't get quite as good results as you expected). This dogma extends beyond the individual attack roll, into encouners and adventures: the PC's might fail, but they'll probably succeed (even if they don't succeed quite how they expected to).
In my experience, success is only sweet if the chance of failure is very great -- if you succeed "against all odds," then you succeed based on your choices and your tactics and your skill, rather than luck. That makes success very sweet indeed, and if you wanted a magic casting mechanic to reflect that, a "failure chance" isn't really the best mechanic to use. It doesn't model tactics or skill at all -- just luck. Instead, perhaps something like an Iron Heroes "charging" mechanic could be in place, or some sort of cascading effect combo chain, or just a ritual mechanic if you don't intend for magic to be combat-viable. Whatever it is, it should be something that allows the player to make choices that can enable the "impossible!" to happen. A dice roll doesn't do that very well, because it's pure luck, and OF COUSE luck is stacked against you. And if this is one of your central character choices (you're spending resources on it), a 90% chance to do nothing for your turn means that 90% of the time, you're a useless sack of meat.
If the chance of failure is only average, or less-than-average (as is the case with attack rolls and saving throws), failure adds no real sweetness. You're failing while everyone around you keeps succeeding, you feel like you're the doofus of the night. If there is a 30% or 20% chance of failure and you get it, it's a negative thing -- you're the only useless sack of meat at the moment. But there's always the comeback -- you're pretty sure next time you'll hit, or the time after that. You only suck for a round, and then you're back into it, and sooner or later, EVERYONE sucks for a round here and there. Furthermore, the luck is slightly in your hands -- the next round, maybe you'll try a slightly different tactic, or attack a different target, or whatnot. If this is one of your central character choices, you're no worse off than anyone else in the party.
In my experience, low-magic-imposed failure chances on magical bang-zoom-pow tends to adopt a "rare yet potent" kind of dogma: you probably have less of a chance to light it off than the melee warrior does to whack the bad guy, but if you do, it's going to BLAST the bad guy.
It sounds fine, but it's the "save-or-die" problem. You fail, not enough happens. You succeed, too much happens. Either way, nobody wins.
An alternate low magic method might be something like a "high cost" kind of dogma: you can BLAST the bad guy, but if you do, you're somehow weaker for it. Perhaps it also takes a lot of character resources to learn to do it in the first place. You pay the cost, a lot happens. You don't pay the cost, you can still do other stuff. Either way, everyone wins.
D&D's higher magic method is a "special effects" kind of dogma: you can whack the bad guy with a magical twinkle dust, or you can whack him with your super battle-axe ninjutsu, but they're going to have effects that are on par with each other, and you can do both about the same number of times.
I totally understand that's not what a low-magic game wants, but that doesn't mean there needs to be a 40-60% chance of "you suck!" in order to evoke that low-magic style.