I had a conversation today with a friend who is a researcher, which put things in an interesting perspective.  The point he made was that just because Hillary out-performs Obama with certain constituencies when she is head to head with him, doesn't mean that she will outperform him vs. John McCain.  Indeed if there is a correlation, it is probably a relatively small one.

Look at it this way.  Let's say you take a population of people and offer them ice cream.  They have two choices - vanilla or chocolate.  Let's for the sake of argument say that more people pick vanilla.

Then, on another day, we offer the same population ice cream again, but this time the choice is vanilla or strawberry.  The fact that people chose vanilla over chocolate gives you almost zero information about whether people will pick vanilla over strawberry.

And... it will also give you no information for predicting whether the same people would pick chocolate over strawberry.

This entire effort on behalf of the Clinton campaign to show that  HRC's stronger performance in one contest is predictive of what her performance would be in another is completely illogical.  Its like saying people who choose vanilla ice cream over chocolate will also choose vanilla over strawberry.