The shake of the forming wire does help align the fibers in cross-direction, but the effectiveness of this is smaller than all the other effects combined, which align the fibers parallel to machine direction. So, overall, the process reduces the alignment to machine direction somewhat, but does not change it enough to make cross-direction dominant. So, everything makes sense: toilet paper is a long band, grain is aligned with machine direction (the long side of the band, perpendicular to perforation), and that's why it tends to tear "the wrong way". It was exactly the same with the paper towel whose picture is shown in the post.