I've been struggling for a week trying to optimize the combined throughput when beam combining with a PBS. With 445 diodes oriented so that their near field patterns looks like === and passing these beams thru a wave plate the result then passes through the cube with negligible loss, but the unrotated beams striking at a right angle don't reflect with low loss to join the transmitted beam, but loose about 30% to wasted transmission. Tipping the cube to reduce the plane of the interface to less than 45 degrees with the first (rotated beam) reduces the loss of the second beam , but only at the expense of loss in the transmitted beam. Its as if the lasers are not well polarized. To confirm that I didn't just fail to line up the lasers with each other on installation and produced a blended polarization, I tried to orient the cube with just one diode and no matter what orientation the cube would not fully reflect or transmit the beam. I tested two cubes one from Stanwax and one from Connecticut lasers and both preformed identically. Any thoughts?