This OSEM placement is just the OPPOSITE of what the proper placement is.
Usually, we want to put them in so that the LED beam is vertical. This makes the OSEM immune to the optic's vertical mode.
The orientation with the horizontal LED beam makes the immunity to the side mode better, but may spoil the vertical.
In reality, neither of these assumptions is quite right. The LED beam doesn't come out straight. That's why Osamu and I found that we have to put in some custom orientations.
Also, the magnet gluings relative to the OSEM bracket centers are not perfectly aligned. So...I am saying that the OSEMs have to be oriented empirically to reduce the couplings which we want to reduce.