This proposal doesn't really do away with the electoral college, but rather causes it to operate differently. That is a good thing, because the electoral college plays a valuable role, but not for the reason you attribute to the Framers.

There is far less variability in political thought among states than within any particular state. You can find conservatives in Massachusetts just as you can find liberals in Utah. So the regional candidate isn't so dire a threat. Incidentally, this fact also mitigates the supposed harm that results from candidates ignoring any particularly safe state: whatever the views held by voters in that state, voters in other states hold them too. The Framers probably would have been happy about that.

The real advantage of the electoral college is in discretization of the vote. When we have a close election, the lawyers and partisans are able to focus on the votes in a few states (Florida, for instance), rather than prompting nationwide recounts. Compare this to the situation in Mexico right now. Obrador hasn't accepted the official result, which is that he barely lost to Calderon. Depending on the rulings of various courts, they may have to physically recount all 41 million votes cast in the election. Every recount they've done so far (each of which has been at a summary level rather than ballot by ballot) has had a slightly different result, so this is clearly a painful process.

The Framers probably wouldn't have used the word "discretization" (if in fact that is a word?), but they did have a sense of how difficult it is to reliably count votes. By setting up a system more tolerant of errors in cardinality, they made accurate (in terms of the chosen rules if not in terms of "absolute democracy") ordinal results more likely. The proposal trades some fault-tolerance for some democracy, by making the vote slightly finer-grained. That is, the smallest increment in an electoral college contest will now be one electoral vote rather than three (the number of votes that very small states get) or more. That's probably OK.

Are near-ties so common that concern for them argues for retaining the electoral college? First, near-ties are the only situation that matters; no one cares about the difference between losing by 40 points and losing by 41 points. Second, as political parties have become more professional, they have become more skilled at slicing and dicing the electorate. Barring external events, they are less and less likely to nominate a truly unelectable candidate like a Mondale or a Dole. (except perhaps in contests against incumbents deemed unbeatable?) A close election in a two-party system is an indication that both parties have done their jobs well. Near-ties aren't going away and so the electoral college shouldn't either.