Hmm, on closer inspection it seems on first approximation that the length minimizing of members in compression is of great importance to frame weight minimization. Stuff buckles in compression at much lower loads than it breaks in tension, and length exacerbates this. ie buckling strength is proportional to 1/Lē.
That would mean focusing on the rear triangle's seat stay and the main triangle's top tube. (Probably the seat tube minorly.)

This would result in a straight angle between both the seat tube and the seat stay, and the seat tube and the top tube. But then you'd get problems in the very shallow rear triangle again.

That actually would be low bike. Much depends on the angle of the seat tube - a lower frame for more angled seat tubes. A straight top tube for a vertical seat tube, the rear triangle would be a compromise.

I gotta check back on this later somewhen, probably have to write some scripts and iterate it...