Alcohol-related fatalities on America's roads fell to a record low last year.

Or did they?

Federal stats on drunken-driving deaths aren't real numbers but educated guesses, based on incomplete reports. Many states - including Florida - do not routinely test and report blood alcohol levels of drivers in fatal crashes.

The oversight skews crash data, the government concedes.

Worse, it puts impaired drivers back on the road, says John Moulden, president of the National Commission Against Drunk Driving.

``The best do-not-pass-go, get-out-of-jail ticket for somebody involved in a drunk driving crash is to get in the ambulance and go to the hospital,'' Moulden says. ``Nationally, we only test 25 percent of the drivers who survive a fatal crash.

``Three-fourths of drivers who live through a crash fatal to somebody else never get tested for alcohol.''

Numbers may be the boring face of drunken driving countermeasures, but they're central to saving lives, Moulden says.

Numbers convict drunken drivers. Numbers lure money for programs. Numbers gauge progress.

Florida's numbers aren't good.

Last year, the state reported test results on just 17 percent of surviving drivers, according to the National Highway Traffic Safety Administration.

That means 1,981 drivers dodged an alcohol test after a fatal crash.

Thirty-five states did better, including neighboring Georgia, which reported 72 percent. Georgia law requires tests of all drivers in fatal or serious crashes - not the case in Florida.

``When only 17 percent are tested, how in the world do you know if you're doing anything to affect drunk driving?'' Moulden says.

``The majority of people, you don't have a clue about.``

Lacking test results, federal analysts considered details such as crash time and driver age, before estimating Florida's alcohol problem.

The shaky numbers bleed into other reports.

Florida earned an ``A-'' this month when Mothers Against Drunk Driving graded states on laws, programs and crash trends.

The rating was based, in part, on a 15.9 percent drop in alcohol-related fatalities from 1993 to 1998 - a calculation that hinges on the accuracy of the federal statistics.

Federal alcohol statistics come with a disclaimer predicting accuracy problems for states with low reporting.

MADD wants mandatory blood alcohol tests in Florida, says Judy Alexander, state executive director.

``Our numbers may not be reflecting exactly how many alcohol-related crashes are occurring that result in death or serious injury,'' Alexander says.

Highway crashes kill roughly 2,800 Floridians annually.

Enough people die annually in U.S. car crashes to wipe out the city of Tampa in just seven years. Nationally, road deaths hover at about 41,000 a year.

It's been that way for two decades, with little fluctuation.

But alcohol now gets the blame for fewer deaths - in Florida, just 33 percent, compared to 47 percent in 1982 - meaning more people die sober.

A changing landscape confuses comparisons: More people are on the road now, inviting more crashes. But more people use seat belts and air bags, escaping greater injury.

The federal government has begun to hold grant money hostage in an effort to persuade states to beef up alcohol testing. By 2001, states must qualify for impaired driving funds by meeting five of seven criteria - one, better testing.

Florida, so far, meets four of the seven standards.

Coaxed by the feds, 50 state experts on highway safety, blood testing, law enforcement, trauma care and statistics gathered in Orlando last month for a workshop dedicated to improving Florida's record.

``Just about everyone seemed to believe there were more tests being conducted than were being represented by those statistics,'' says Al Roop, a law enforcement trainer who mediated the workshop.

Roger Doherty of the Florida Department of Transportation points to one example:

Florida medical examiners conduct autopsies on victims of violence, including crash victims. They checked for blood alcohol and drugs in 92 percent of cases where a driver died, Doherty says.

Yet a federal report shows test results on just 56.5 percent.

``We have to find ways to make sure the testing that's done doesn't fall through the cracks,'' Doherty says.

Police check surviving drivers less consistently.

Agency policies differ.

Florida law requires that police have ``probable cause'' to believe a driver was impaired. In the adrenaline-charged scene of a crash, a driver well tolerant of alcohol may not show signs of being drunk.

``The individual law enforcement agencies, some would argue and say we have mandatory testing,'' MADD's Alexander says. ``That's only because that particular agency has adopted that kind of policy within their department.''

Increased testing could logjam the state's already clogged crime labs, Roop says.

As Moulden sees it, increased testing might ultimately bring safer roads.

``If you don't test,'' he says, ``you don't have a clue of whether anything is getting better or not.

``If you do a better job of testing, you identify drinking drivers.''