A detailed analysis of delegate math leading up to the Democreatic National Convention by Chris Bowers is here.
The one significant flaw in the analysis is its scenario number 5 in which a Michigan delegation is seated on the basis of the January primary. In that scenario while some delegates from Michigan would be "uncommitted" those delegates would in essence be "anti-Clinton" delegates as Clinton supporters presumably voted for Clinton rather than uncommitted.
The bottom line is that it will take a truly remarkable set of events between now and the Convention, in which Clinton receives very favorable resolutions of the Michigan and Florida situations and repeatedly outperforms expected results based upon polls and prior contests for Clinton to become the Democratic nominee.
The other scenario in which Clinton wins is one in which the superdelegates monolithically all choose to buck the pledged delegate count, something which would create something close to constitutional crisis for the party. It is permitted, but an overwhelming disregard for primary voters and caucus attendees would likely wound the process and the nominee. It is not believed to be very likely to happen.
So Obama v. McCain is much more likely than Clinton v. McCain at this juncture.