There is no way I know of to correct alignment of completed anaglyphs after they're rotated.  Therefore, the only solution I can think of consists of not using completed anaglyphs to form the cube but instead, use 2D images.  Wh en the cube is given depth through converting it, the individual 2D images must be converted along with the process.  This will result in much more effort by effectively converting 4 images whereas before, only the cube needed to be given depth through conversion.