Where are people getting this idea that the spell is like Blur? That the images are all in your square?
Because otherwise no one would be fooled? I mean, suppose the mage moves, then (for whatever reason) stops in front of the fighter, then casts mirror image, making a whole lot of images spring up on either side of himself/herself. Why wouldn't the fighter just tock the mage standing right there (after all, s/he knows, presumably, that the mage can't move any more that round).
The situation can be made more puzzling if there are pit traps either side of the square the mage is in - let's suppose that s/he flew in over the top of them and then landed in the one square that is safe (X = pit, O = empty square, w = wizard, f = figther):
XXXwXXX
OOOfOOO
In this situation, if the mage creates images that hover over the trapped squares the warrior would just ignore them, knowing that they must be fakes because if they were the real mage then s/he would be falling down a pit.
Or here is another examle. The wizard casts mirror image with the following spacing of images (hoping to fool a whole lot of archers 50' away), and there is an invisible fighter adjacent whom the wizard is ignorant of, and who is unable to move out of his/he square - perhaps due to engtanglment, or geas, or whatever; M = image):
MOMOwOMOM
OOOO f OOOO
In this example, the fighter cannot reach any target but the wizard, so why is there random determination of whether the wizard or an image is hit? Uness you think that the spell also gives the wizard bonus movement, so that s/he can somehow teleport to one of the squares an image is in.
But it's that "know which square the caster is in" part that screws up that idea. That is exactly what the spell is designed to deny you.
I've given examples where this doesn't work, unless you think the spell gives the caster bonus movement/teleportation.
To do it truly correct, they should replace the figure with a marker, and then place a number of identical markers around it. Mark the bottom of the real one. Then shuffle the markers so that nobody knows which is which.
Where does the spell say that the caster gets to move around like this? And what if the caster is standing on a pillar of rock in the middle of an abyss? Or even my example above, where s/he has pits on either side of him/her?
Okay, so you spend your round doing a Search check to tell which one is the caster, drawing your AoO in the process.
The caster now get's their action, which includes casting a spell and moving. And when they move, the crowd of images shuffles so that you lose track of the real caster. It's in the spell description.
You've just wasted your round.
I already explained why your round wouldn't be a waste. First, if the caster tries to cast another spell you know where s/he is for OA purposes. Second, you can tell your friends which image to attack, so they can take down the mage (which seems more efficient to me than focus firing the images).
Because the spell says so. An attack is aimed, and so you can tell if the target you aimed at is real. AoEs aren't aimed. No specific target, no specific reaction/result.
I know the rules. I'm querying the fiction. A pebble thrown at an image (attack resolved via a to hit roll) can pop it. A war machine rock hurled at the whole area (AoE resolved via Reflex save) cannot pop any of the images it lands on. How do the images know that the pebble was aimed at one target but the machine-hurled rock was aimed at a whole area? Can the spell read minds too?
I think the issue is that this interpretation (which I also favor) is seemingly contradicted by the line of "Generally, roll randomly to see whether the selected target is real or a figment."
I think there are issues with the favoured interpretation as soon as the terrain is anything more than a flat, featureless area. Another possibility I haven't canvassed is this one: the caster is standing in a square at the centre of a T-junction. The images are spread out either side on the top of the "T". An archer fires from 30' away down the stem of the T. If the images are in squares either side of the caster, the archer can't see them, hence can't aim at them, hence has no miss chance against the caster.
Or if, instead, we say the images are in the caster's square or in the squares between the caster and the archer, why does the arrow not just travel through them as it would any other figment, and hit the caster?
I don't get it. I especially don't get how it is meant to fit with the ostensibly "simulationist" tenor of 3E that I frequently see mentioned as a feature of that system.