I've been wondering if the bar even has to be by the TV. Even with the sensors surrounding the TV, the system still needs some way to know where the screen is in relation to them. Unless the sensor bar has to be placed at the bottom corners of the screen exactly, which is unlikely if it's a solid bar instead of two separate units as originally shown, then a one-time calibration in the system menu would seem to be necessary. It would be as simple as touching the remote to the screen's corners and pusing a button, and would only have to be repeated if the bar or TV moved. To make moving the bar less likely, it might be a good idea to put it somewhere where it wouldn't be nudged accidentally, which wouldn't be right next to the TV. Attaching it to the wall would be more foolproof, or it could at least be put behind the TV with the dust bunnies. Since it's presumably used to triangulate/trilaterate the remote's relative position to the sensor bar, it could be anywhere within range, in any orientation, and the math could be done to figure out the remote's relative position to the TV just as easily. It might affect accuracy a little, especially if it were in line with the remote.
I haven't done the math to see exactly how possible all this is, but it seems like there are enough points for 3D positional calculations. Points A and B are in the sensor bar, and because it's one piece, they are a known distance from each other. Points C, D, and E are in the remote, and likewise a known distance from each other. We can use the timing of radio waves to get the distance from the points in the remote to the points in the sensor bar, which should be enough information to calculate the exact position of the remote (including rotation if the points in the remote form a triangle) relative to the sensor bar.
Does anybody see a problem with this?