Floors are a different thing, but we don't spot treat floors. We treat the entire area in front of a speaker. I've never seen anyone put down a 2'x2' carpet exactly in the first reflection point, and this is kind of what i mean.
Eric, I have a dedicated room. I made 2'x4'x4" OC 703 panels that cover first reflection points on both the floor and ceiling. The rest of my floor is untreated. I realize this is not common, but I have a dedicated room and I am the only listener, so I can do what I want. Both ceiling and floor "treatments" have very little influence on imaging in comparison to my side wall "treatment." They certainly offer other benefits. which more than justify their use. I have considered modifying my ceiling panels in order to add forward facing reflection to the existing absorption, but have not done so yet.
As I recall, my measurements show that the floor and ceiling reflections are delayed about 4-8 ms, whereas the side wall reflections are more in the 6-12 ms range. Without treatment, the ceiling and floor reflections are louder than those from the side walls. With treatment, the side wall reflections are substantially reduced, whereas the ceiling and floor reflections remain relatively high. To a certain extent, I think this is indicative of how much more effective reflection is compared to absorption.
From my perspective, sidewall 1st reflection point treatment is conceptually sound, and in my room, theory has been supported empirically. Imaging and localization in my room is superb. It is the best by far that I have ever heard in a narrow room. If you have experienced something different, perhaps that is because you are attempting to "treat" your sidewalls using panels that are not up to the task. Again, I'm going to suggest that before you discount the concept, you evaluate the approach articulated at least twice by Duke and several times by me. Rip a 4x8 sheet of half inch plywood into 2 x 8 sheets and place them along the side wall behind the first reflection points angled so as to reflect sound back to the front of the room and away from the MLP. This is enough to reflect the wavelengths that are important to localization clues, and it turns early reflections into late ones. It is a cheap and easy experiment. Keep in mind that the first reflection "point" is far from a point, because the sound coming from your tweeter and midrange will spread out as it travels rather than proceed as a laser like ray. So you will need to experiment with placement and angle.