Attempting to put hard-and-fast ranges on which things are best at which levels is a fool’s errand; there’s too much variety in what a given level “means” for that to be terribly meaningful. You are also making several classic mistakes in your analysis, most notable of which is assuming that the game at high levels is going to be much of anything like the game at low levels.
Why AC is good early
Nothing particularly “changes” for AC at 10th level, or any other level for that matter. Early investment in AC is good because there are a number of sources that offer a big chunk of AC for very little cost, that you can and should get at very early levels:
Up to at least +7 to AC from armor and shield is usually trivial. A +1 chain shirt and +1 buckler are both usable by almost everyone, and affordable at appropriate levels. Since they’re desirable platforms for other enhancements, there’s also no worry about “waste” later on, even if you stop investing in AC.
Some builds may go for heavier armor and have higher numbers here, but that becomes more costly so it’s harder to generalize about.
Dexterity is good for a lot of things (it is second only to Constitution for its universal desirability), so a lot of people will get some of that, too. Ability scores start high and scale slowly, so most characters have most or all of the Dexterity they’ll ever have right from 1st. (Even if you focus on Dexterity, your 1st-level Dexterity is going to be “most” of your Dexterity bonus probably until a wish or manual of quickness of action come into play.)
These have nothing to do with level; they never stop being good, because they provide other benefits. In the case of the armor and shield, they’re also extremely inexpensive. No matter what level you are, you want to have these things. The only reason they’re associated with low levels is because you can afford them at low levels—and since you can, you probably should, because they’re good.
The key thing is that after you have your initial Dexterity score and +1 armor and +1 shield (if you’re using one), getting more AC becomes very expensive: you’re dealing with quadratically-scaling costs. A ring of protection +1 isn’t too bad, but higher numbers are very costly. An amulet of natural armor +1 probably isn’t even worth it, since it doesn’t apply against touch attacks; likewise, if you were allowed to skip the +1 enhancement on your armor and shield, you probably would. And then higher values of those become even more expensive. Increasing Dexterity can obviously be worthwhile, but it comes slowly and expensively and if you’re not Dex-focused, you probably won’t be able to afford much of it, at least not until quite high level.
Note that all of this is “in a vacuum,” so to speak. For it to be meaningful that these upgrades cost so much, there has to be something else you could spend the money on that would be more meaningful. This is where miss chances are usually brought up: they’re a fairly direct, apples-to-apples comparison with AC, because they protect against the same things. And indeed, your chance to be hit by an attack is reduced much more cost-effectively by miss chances than it is by AC once the “freebies” to AC are exhausted and miss chances become more available. But even without miss chances, the rationale would still hold true: there are tons of things that you can get with the enormous amount of gold that AC upgrades cost. Cataloguing everything you could have is far beyond the scope of this answer (or, really, any Stack Exchange answer; that is the subject of a full handbook, and several on this subject already exist so I’m not going to write another one). But it needs to be pointed out that this is the case regardless of miss chances.
Why optimizing high levels doesn’t really care about AC or miss chances
Optimized high-level D&D 3.5e is generally rocket tag: if an enemy can target you with something you aren’t immune to, you’re as good as dead, or worse. Your defenses are, primarily, immunities, and simply not being where someone can target you. But even those aren’t really your defense here: while it’s possible to manage true immunity to everything, from a practical standpoint you’re probably not going to be allowed to play such a character, and as long as you have vulnerabilities, something will find them. And as long as you want to do things and affect the world, you have to be somewhere that you might be targeted by one of those things. So the real answer is “the best defense is a good offense.” They can’t target your vulnerability if they’re already dead. Hence, rocket tag.
Which means that your entire image of what D&D combat looks like is wrong, at these levels and this much optimization. No one is standing toe-to-toe with anything; they’re using divination, stealth, teleportation, and so on to kill targets without giving that target any chance to respond. Contingency is the most important effect in play simply because it might give someone a chance to respond always, no matter what (depending on how clever a spellcaster is with their trigger—and how clever they’re allowed to be). A sufficiently-paranoid spellcaster is thus basically unkillable.
And the answer for martial characters is, they don’t get to play that game. While spellcasters were moving past “combat” altogether, martials were just getting incrementally better at combat. Once you don’t really do “combat” at all anymore, those skills are kind of pointless; at best, a buffed-up fighter teleported on top of an enemy might be an effective nuke instead of directly targeting them with a spell.
This is, to most players and certainly by the descriptions of the game in the books, dysfunctional. The game isn’t supposed to be like this. With a healthy dose of gentlefolk’s agreements and sufficient system mastery shared among everyone to avoid it, you can hold it off, though it gets harder and harder as you gain levels. What exactly that looks like—and thus how to best optimize for that environment—is going to vary a lot from table to table, though, so it’s impossible to really give much general advice on what would be ideal for martial characters.
Conclusion
High-level D&D 3.5e is broken, at least according to the books and what most people expect. Your options, effectively, are to avoid it, ignore it, or embrace it:
“Avoiding it” means simply not playing at high levels. This is, by a vast margin, the most popular answer: the overwhelming majority of campaigns start at very low levels, and almost none of them last long enough to reach these levels. This isn’t usually a conscious decision, mind: it “just happens” as far as the participants are concerned. (The fact that the game just gets harder to run and play at higher levels probably does influence this, but for most groups this is probably not consciously noted.)
Of course, some do make very conscious decisions about this: that’s what E6 is. I, for one, heartily recommend that.
“Ignoring it” means, basically, ignoring the campaign-warping effects available and continuing to play the game as if it was still low-level. For those few non-optimizing groups that reach high levels, this is usually what they are doing, if unintentionally. If clerics just cast cure and heal and wizards just cast fireball and lightning bolt, and monsters just cater to martials and give them the fight they want, it’s possible to just not run into the various problems the system has at high levels.
This is basically incompatible with optimization of the written system because the advantages offered by the campaign-warping spells are so immense; of course it would be optimal to leverage them. Likewise, if opponents were thinking rationally and optimally, they would just prevent fights with martials from ever happening, which many monsters are quite capable of doing.
Even without dedicated optimization, the entire thing is fragile: all it takes is for someone to notice some of the more-optimal options they’ve been ignoring to start breaking the game down. The higher-level you get, the harder this gets to ignore. This—I suspect—is part of why more casual games rarely ever reach high levels, because stuff comes up that makes maintaining the illusion harder, and things get more complicated and slower: the game just becomes less fun, slowing down without noticeably gaining new fun new dimensions to explore.
The only way around this is to liberally use houserules and general agreements about what is or is not appreciated at the table, and keeping within those limits. This, ironically, requires immense system mastery, because you’re trying to optimize while at the same time trying not to optimize “too much” and trying to march right up to that line without crossing it is hard.
“Embracing it” means, probably, everyone is playing a spellcaster. Anyone who is not, is trying to synergize specifically with the spellcasting their allies are doing, optimizing for scry ‘n’ die tactics or whatever. In the end, this game is very little like “D&D” and it becomes much more about intrigue and politics. The actual rule system of D&D 3.5e ends up mattering fairly little, for all that everyone involved is heavily invested in that system, because there’s not really any “system” involved in intrigue and politics.
But if you embrace it, this question becomes close to moot.
