One last thing that seems curious to me is that the 240p optimal modes seem to operate around a different set of ideas about how aspect ratio should be handled, at lest for a 240p source.
Yes, this is by design. The purpose of optimized mode is to accurately digitize the source on a dot-by-dot basis, and output them as-if square pixels. As the source material in almost all cases does NOT have square pixels, the output image is a compromise in regards to image aspect ratio.
So 320×240 optimized output will normally be too wide, and 256×240 will be too skinny (8:7 mode) or too wide (4:3 mode, and to varying degree depending on line multiplication mode).
It can be corrected in a second step by your display (if it has advanced controls enough, which most won’t), a secondary scaler, or by capture software. (see another currently active thread on this subject)