>and that could be used to detwin the DANOs directly:
DANOtrue(h1) = ((1-tf)*DANOTw(h1) -tf*DANOTw(h2)) / (1-2tf)
DANOtrue(h2) = ((1-tf)*DANOTw(h2) -tf*DANOTw(h1)) / (1-2tf)
This would be problematic, since many of the signs of DanoTrue would not be
known per the Keller Theorem just posted. This would not be a problem, however,
when detwinning I+/I-.
A corollary: the twin fraction at which one reflection from a twin-related pair
will flip sign is a function of the ratio (Dano/Danotwinrelated); higher ratios
flip at lower twin fractions. All Dano twin-pairs with ratios < -2, for
example, will have flipped sign with twin fractions > ~0.33. Working right now
on pinning down the right function.
JPK
