My statement was true in 2006, though - less so than 2004 obviously.
Ok. I think I would need more datapoints (as we have with several years of N.J. and, to a lesser extent, Virginia polling wildly in the wrong direction long before Election Day) before we could call this a law.
In 2004, everyone acknowledged that polling in Florida was disastrous because the state had been hit with four hurricanes. Many articles said "no one knows what's going on there" when reporting on it. What was going on, that didn't get much reported or remarked on until after Katrina, was that FEMA was firing cash at Floridians like a firehose with lots of fraud and excessive benefits that people hadn't asked for. Among other things. At the same time, Florida's economy was still booming, and if anything employment was up as a result of the reconstruction boom. Jewish retirees from the tri-State area unexpectedly swung from the Democrats there as their friends and children did in New York and New Jersey. Finally, the profile of many parts of Florida matched the general profile of strong Bush regions--fast growing suburbs and exurbs with young families and weak ties to government, with a mix of religion and a dearth of union laborers.
Florida is a state where things can change quickly. The boom turns to bust. A mass migration turns to a trickle. The cost of living jumps. Communities change their ethnic make-up. I would hesitate to extrapolate from 2004 to future presidential elections as a "law," particularly when we can't extrapolate backwards with clarity, and I would also be reluctant to read any race with Katharine Harris in it as if it were a model for anything else. She did better because a lot of Republicans skeptical of her or outright embarrassed decided to vote for her in the end. This is not the kind of dynamic that moves a McCain from 48% to 51% or even from 51% to 54%.
I would say that we don't know what's going to happen in Florida, that McCain has an advantage vis a vis the national average, but that we do not have evidence to support the idea that Florida has a built-in Republican edge that is invisible to polls aside from M-D but will come out on Election Day. It's too much of a just-so-story from one election that has too many other possible explanations and not enough datapoints.