The reason "M" is better is because overall font metrics include EVERY glyph, even ones that extend an unusual amount outside the "norm" for the font. I suppose you could use "font size" to come up with something relative to the baseline, but that will require some tweaking to get results that look good across a variety of fonts.
Also, "M" is most like what "everyone's existing schematics" is already using, and is a decent representative of "average glyph size" even if the "M" (or other iso-latin-1 glyphs) doesn't occur anywhere in the string.
The reason "M" is better is because overall font metrics include EVERY glyph, even ones that extend an unusual amount outside the "norm" for the font. I suppose you could use "font size" to come up with something relative to the baseline, but that will require some tweaking to get results that look good across a variety of fonts.
Also, "M" is most like what "everyone's existing schematics" is already using, and is a decent representative of "average glyph size" even if the "M" (or other iso-latin-1 glyphs) doesn't occur anywhere in the string.