Your math and logic checks out. Balance-wise, that's much trickier.
First, let's address stacking. Compare the price of bracers of armor +5, which cost 25,000 gp, with an amulet of natural armor +5, which costs 50,000 gp. They both do the same exact thing, but one—the bracers—doesn't stack with normal armor, whereas the amulet does. Because it's easier to stack with more common sources of AC, the amulet of natural armor costs twice as much.
The primary way of reducing arcane spell failure, weight, and Dexterity limit is through armor material. That's important for two reasons: first, armor material scales with armor weight class. Mithral might cost 4,000 for chainmail, but it costs 9,000 for full plate. Something like entropium (from the Arms and Equipment Guide), which also reduces arcane spell failure, armor check penalty, and improves maximum Dexterity bonus, scales from 2,000 gp for medium armor to 8,000 gp for heavy armor. Blended quartz (also from A&EG) costs 5,000 for medium and 10,000 for heavy, and it just reduces arcane spell failure.
Second, as a magical armor ability, celestial would, in theory, stack with armor special materials. Just like the amulet of natural armor, that means it should probably cost a premium.
Now let's look at armor enhancements. The twilight armor enhancement (found in Player's Handbook II) reduces arcane spell failure by 10%, and is a +1 enhancement. The nimbleness armor enhancement (in the Magic Item Compendium) increases maximum Dexterity bonus by 2 and decreases armor check penalty by 1, and is also a +1 enhancement. Celestial is clearly stronger than either of those. Being a +2 enhancement might be reasonable. That would mean celestial armor, the item in the DMG, is much cheaper than the sum of its parts, but that's not necessarily a problem.
We can also quickly check armor properties that cost around 5,000 gp to check your initial estimate, and we find quickness, which adds 5 ft. of move speed, and then properties like bouyant (+2 to swim checks, no ACP on swim checks), landing (ignore 60 ft. of fall damage), or gilled (water breathing). On the stronger side of that price range are things like 3/day haste for 1 round from the speed enhancement. If these all seem weaker than celestial, that's because they are. 5,000 gp might be a mathematically correct price derived from the base item, but it's clearly very cheap for the effect.
Personally, if I were to allow an isolated celestial enhancement, I'd start with scaling it to armor category like a special material. Blended quartz doubles in price, entropium quadruples in price; I'd go somewhere in between and triple it. 5,000 for medium armor, sure, or 15,000 for heavy. This is more expensive than the aforementioned materials, but it's also better in nearly every way, and given that it stacks with special materials, a bit of a premium is sensible. Light armor could be 3,000 or something; celestial is pretty weak on light armor. (I'd also apply the price before factoring in mithral lowering the armor category; celestial mithral fullplate would still cost the +15,000.)
But that's in a vacuum, and balance questions (which this certainly is) are never really in a vacuum. So let's look at whom such a property most benefits: gishes, and to a lesser extent thief- or archer-types. If your campaign has someone who's got, say, 13 levels of scout and wants to wear heavy armor without it impacting his ability to shoot things too much, then maybe the 5,000 gp price point is reasonable. Even someone like a duskblade, or just a run-of-the-mill eldritch knight, probably isn't outshining the rest of the party, higher AC or no. On the other hand, if you've got a reasonably-optimized abjurant champion, then 17,000 gp for a suit of +1 celestial full plate seems reasonable. Alternatively, leaving such a suit or armor as loot somewhere might be a better approach, depending on the party.
Ultimately, 5,000 gp is much, much cheaper than equivalent effects, and celestial is much stronger than +5,000 gp armor abilities, but depending on who's benefitting, that's not necessarily a problem, since Dexterity-based characters and poorly-optimized martial/mage types could probably both use some help.