It’s been said “If you torture the data long enough it will say what you want.”
Taking all these “second choice” polls and their respective percentages of who would take who IF Donald is taken out and then somehow combining those percentages (curiously not mentioned as what percentage of the entire slate of candidates’ supporters) to reveal, voila!, some concoction that says, “Rubio or Cruz beats Hillary” is absurd.
This wasn’t a specific poll for that; it is a mishmash of wishful extrapolation of orthogonal data from RCP averages that leaves out some very basic obstacles - like how exactly Trump magically disappears from the entire data set.
But Donald isn’t winning against Hillary in a General Election, in any of the polls that I read on Real Clear Politics (unless I missed it).