Yes, 1 AC at 1st level and 3 AC at 20th are not nothing, but the flip side is that -2 whenever bladesong is not active is also not nothing. The alert feat solves only one of the problems of that -2 AC, and only in games where feats are allowed, and the player has to play variant human to get it without delaying ASIs, which in turn lowers AC.
You're right about armor types; I was thinking that mage armor was 12 + Dex. Brain fart. Edit: Though it has just occurred to me that a fighter wearing armor instead of using mage armor could take the defense fighting style to gain +1 AC. That would give plate an advantage at the highest levels, and perhaps other armors earlier depending on how much money the PCs have to throw around. /Edit
By delaying ASIs, you negate that +1 AC bonus you got from the bladesinger dip from 4th through 8th levels on the eldritch knight, and that's just one of the drawbacks of having lower Dex. Compared to your build, a single-class eldritch knight in that range has +3 AC generally, the same AC when bladesong is active, +1 to hit, and +1 to dex saves. At 6th level, it has an additional +1 AC, +1 to hit, and +1 to Dex saves, and at 8th it has +1 Con, +1 Int or a feat. Your build provides +1 AC part time until 4th level and then doesn't do so again until 9th. That's a pretty big problem.
The build might work better with rogue, but so do most builds. Rogue is up there with warlock among the best multiclassing options.
What you say about blur is correct, up to the point that you consider. If an enemy has a 50% chance to hit you, blur will reduce it to 25%, which is equivalent to 5 AC. However, if you have an AC of 26, an enemy needs to have a +15 to have a 50% chance to hit you. You're talking about reaching that AC at 14th level, while only monsters around CR 21 and higher have that kind of to-hit mod. Most of the time, you are going to be facing enemies with a lower chance of hitting you. Around CR 14, you're looking at the neighborhood of +10, around a 25% chance to hit. So, with blur, that becomes a 6.25% chance to hit, your gain equivalent to 3.75 AC. Against swarms of mooks with only a 5% chance to hit you in the first place, blur is worth less than 1 AC. This is why I say that blur loses value at the kind of AC you are talking about.
Mirror image isn't spectacular at those AC levels, either, but it has the advantage of not being a concentration spell, hence having no chance to break if you're caught in a random dragon breath or cone of cold, or if you're hit by a status effect, or if you cast a concentration spell, or if you ready a spell. In fact, how are you even maintaining 26 AC with blur, if part of that AC is coming from shield of faith? Mirror image is also front loaded, such that its effective AC contribution is often higher than blur's for the first X attacks, depending on the to-hit mod of the attacker. Personally, though, I'd not cast either.
The comparisons you need to make are to single-classed characters and to multiclassed characters who take the wizard and cleric dips at different points. Is it really beneficial to an eldritch knight to get bladesong before having Dex maxed? Or, for that matter, before having Int maxed? Is it worth delaying extra attack? Uncanny dodge? ASIs and feats? Class archetype?
Suppose I started eldritch knight with 16 Dex, 16 Con, and 12 Int, maxed Dex at 6th, bumped up Int at 8th, and then dipped into wizard and ditched the shield at 9th. You need to make the comparison and discuss the differences in HP, AC, saves, spell levels, spell slots, play styles, etc., and then you need to make the case for your build in light of those facts. You need to demonstrate not only that your build actually does what you want it to, but that what is does is worthwhile, and that the investment in Int is worth the cost in Con for any character whose main class is not wizard. And you need to show that it works at every level, not just 20th.