So a single atomic electron that gets excited in this way indeed does not have this property, and indeed a vast number of them can do this in parallel without functioning this way: this is part of the classical explanation for why the sky is blue.
Furthermore it matters that the surface is flat: if you cut little parallel lines in the surface then you get a spectrometer; this is part of the explanation for why you see rainbows in the "data track" of a CD or DVD.
Furthermore you see the same reflection when you analyze things like reflection from a glass, even though there is also a transmitted wave.
Our modern understanding is that a photon is able to sense every path that it could possibly take from A to B. Each path can be thought of as a little arrow rotating in 2D at a constant rate with respect to time (the photon's "frequency"). We add up all of these little arrows by connecting them tip-to-tail and then ask how far away the final point is from the initial point, which is a measure of the probability that the photon goes in that direction.
Now if there is a path, and there are other possible paths "nearby" that one, like on a mirror, we have to consider if they take a different time or not. If they all have a different time then each arrow is tilted a little more relative to the last one, and as you add them together, they describe something very circle-like and it will not get far away from the initial point. But if the path is either the fastest or slowest path from point A to point B, then something magical happens: all of the nearby paths take about the same time, so all of the arrows line up, so you get a very large probability that the object moves in that way.
This is known as "Fermat's law of refraction," and it says that light seems to magically find the shortest-time path between two points. And a direct consequence is that the angle of incidence equals the angle of reflection.
