Your math is correct assuming the rendered anamorphic image is 640px wide inside the 720px active horizontal width.
However, I’ve observed that, when set to widescreen mode, the Wii actually seems to render content across the full 720 active pixels, or at least much closer to the full 720 than to 640 (some content has more blank padding than others). In contrast, when the Wii is set to 4:3 mode, it instead renders 640 pixels wide as expected (though at the same 858 total dots/line as in widescreen mode).
I invite others to confirm/refute the above stated observations, but assuming the unstretched anamorphic image (not including horizontal blanks) is indeed 720×480, the desired horizontal scaling (to end at 853.333×480 nonsquare pixels) is 32/27 = 1.185185…, whereas for 640×480 the stretch factor would be 4/3 = 1.333…
Regardless of it being a 4/3 or 32/27 scaling that’s needed to properly display the anamorphic image at 16:9, the OSSC’s means of doing the “stretching” should ideally be via horizontal integer scaling as opposed to interpolation. Basically just like how the OSSC already can “stretch” a 256×240 image to a 4:3 ratio via a 5x horizontal scaling with a 4x vertical scaling. That kind of solution for this situation with the Wii would require at a minimum doing line3x for the 480p Wii output (1440p), which isn’t an option on the OSSC hardware but maybe its successor?