The events surrounding Hurricane Sandy increase the chances of a scenario in which Obama wins the presidency but loses the popular vote. Many argue that that would undermine the legitimacy of an Obama win and the nature of his mandate. I found those arguments unpersuasive when posed against Bush and I find them unpersuasive now. Candidates and the voters behave in response to the parameters of the election. The fact that Obama is likely to win California and New York while Romney is likely to win Texas affects turnout in those states. Why go to the polls if you know your guy is going to win or lose? Similarly, it affects the way candidates campaign. If winning the popular vote were the relevant consideration, candidates wouldn't be spending all this time in Ohio at the price of their other sources of potential supporters.
Measuring candidates by popular vote when they've been aiming for electoral college votes is like changing the baseball criterion for success to net runs over the season after teams having been simply pursuing wins. Teams would object that had they been told that net runs had been the relevant criterion they would have followed much different strategies in games in which they were far ahead or far behind. We can't define success by one criterion and then measure success by a very different one particularly when the metric's definition has such a direct effect on the nature of performance.