Bob Barr certainly could be a threat to John McCain, but it's worth noting that third-party candidates on the right have not done well in previous elections. Pat Buchanan, who is better known than Barr and did much better in the Republican primaries than Ron Paul, got 0.42 percent of the vote as the Reform Party nominee in 2000. Paul did about the same as the Libertarian Party nominee in 1988, just four years out of the House. The U.S. Taxpayers/Constitution Party has never cracked 200,000 votes. John Schmitz, a sitting Republican congressman, won 1 million votes against Richard Nixon in 1972, much better than John Ashbrook did in the primaries. That didn't stop Nixon from winning a 49-state landslide with over 60 percent of the popular vote. The Libertarians' best showing in history, with Ed Clark in 1980, similarly failed to put a dent into Ronald Reagan. Even if you go back to George Wallace's 13.5 percent and 46 electoral votes in 1968, Nixon still won the presidency.
Long before liberals worried about the Nader effect, conservatives blamed disgruntled Republicans for handing the White House to Bill Clinton in 1992 and 1996 by voting for Ross Perot. Perot wasn't a conservative, but his campaign had some center-right appeal. Whether Perot actually tipped those elections is debatable, though there is strong evidence he flipped some states away from George H.W. Bush in '92. I'm not trying to talk down Barr's candidacy -- he is absolutely the right candidate to exploit McCain's weaknesses with the right and could cost him the election -- but the right's appetite for third parties starts strong in early polling and weakens as November approaches.
Another thought: Barr could get close to a million votes just by consolidating antiwar, libertarian, and conservative third-party voters without taking a single vote from McCain's hide. So a replay of Reagan-Clark 1980 may not be likely, but neither is it impossible.