I think I may have just figured it out, but not sure if I can describe it.
I think it is skewed because I am not folding with perpendicular lines. But since I am rotating with the exact opposite angles, I thought it would balance out. I shall try it with a piece of paper.
That would be correct if the hinges were parallel, but they are not. You can construct the geometry as a series of mirror operations and if you then measure the fold angles the ridges and valleys are approx 2.5 degrees different. Here you can see how the discrepancy builds:
Having checked that your first two triangles are both right and identically sized, and assuming that the first ridge is deliberately horizontal,I created three point planes from them, using the common edge ends as the first two points:
I used these two planes to create the bisector plane and, the ridge being horizontal, as the edge of the bisector plane can be confirmed to be vertical I know that the two triangles are correctly aligned:
I then needed a new vertical plane along the valley for the next mirror (for the panel array to remain level, all the mirror planes must be vertical). As ridge and valley meet at the edge of my first bisector, I used its top and bottom corner as the first two points in a three point plane, with the bottom of the valley as the third point. This then allowed me to create the third and fourth triangles by mirroring the first two (note that I’ve drawn the plane here for clarity but I could simply have used the three points in the mirror command):
