TCP orientation and structured light scanner perceived orientation disagree

So I’ve run into an interesting issue. I have a structured light scanner on the flange of my UR arm. When I have the arm return to a TCP orientation state, the robot believes the tooltip is offset by z=90 degrees, and x and y by 15 degrees. My super high fidelity structured light scanner, however, measures an angle difference of 19 degrees in y. What could be causing this disagreement? The structured light scanner’s results have been externally verified.