The CF samples are very interesting. I would expect the fibers would be oriented parallel to the extrusion which is consistent with the results.
The flat vs. edge differences could possibly be attributed to perimeters vs. infill. The edge parts are all perimeters (or anyway a much higher fraction) while the flat parts have infill at a 45 degree angle.
I would guess that a flat part with a very high number of perimeters could have the same stiffness as the edge part. If so it would confirm that the fill orientation is the reason the flat part is less stiff.
Slots in a part can force the orientation of the perimeters, which could make a significant difference for CF parts (besides the ordinary benefit of extra density in selected areas). That could make for interesting designs specific to CF.