Robert Liles, 41, is fairly sure he ran Sunday's Honolulu Marathon side-by-side with Skyler DeWitt. After all, the two crossed the finish line together in 6 hours and 20 minutes. But according to the official record of the event, DeWitt wasn't even at the race.

With two solid marathons under his belt, Brent Martin, 24, knew the wise thing to do was run conservatively in the first few miles then pick up the pace later on. But even the top runners in the field couldn't have run the first 6.2 miles in 2:38:56 and the last 20 in less than an hour, as Martin's online results claim.

Pei Shin, 29, was looking at the official race clock as he dashed through the finish line. He saw 6:11:something — and his actual time should have been even faster given that he didn't cross the start line exactly when the opening gun went off — but his official marathon finisher's certificate reads 6:22:40.

The answer to such riddles lies in the unfortunate confluence of bad weather and fragile technology, but Honolulu Marathon Association president Jim Barahal said the bottom line is simple:

"We screwed up."

Hundreds, perhaps thousands, of runners in the 35th Honolulu Marathon had their split and finish times incorrectly or incompletely recorded when the event's new timing system — administered by Michigan-based SAI Timing and Tracking — short-circuited in the heavy rain that preceded Sunday's race. Worse, as many as 3,500 runners may not have been recorded at all.

"That's the doomsday scenario," Barahal said.

Barahal acknowledged that the marathon association could be faced with lawsuits from angry runners — "Obviously there is the potential for natural repercussions when there's a major foul-up," he said — but the bigger problem could be whether the incident sours the marathon's relationships with its sponsors.

"I don't think it's going that well now," Barahal said. "We have a reputation for doing a great job, and this was a fairly major screw-up.

"I think people understand that we don't do the timing ourselves, but they trust us to make the right decision. Clearly this is not the right decision, and I made it."

TRYING A NEW SYSTEM

The Honolulu Marathon was among the first major marathons to use microchip timing to collect split and finish times, introducing the ChampionShip system in 2000.

The system was effective but costly. Because of the size of the field and the international make-up of the runners, Barahal said it was impossible to collect the plastic ChampionShip lace-ons after each race for reuse. He estimated the cost of offering single-use chips at $250,000 a year. The large timing mats that read the chips as runners passed were also bulky and difficult to set up.

This year, the marathon elected to try a new system developed by SAI that offered an estimated cost savings of $300,000 over five years and was purported to be more flexible in its applications. Had the system worked, Barahal said he was planning on having timing stations at each mile marker and real-time information transmission along designated points at next year's marathon.

Barahal said the race's long relationship with David Simms, who broke away from ChampionShip to start SAI, was a major factor in the decision.

"We didn't just pick this guy out of the yellow pages or do a Google search for timing companies," Barahal said. "He's an honorable, top-notch guy who's always given us phenomenal service. When he told us this was what he was doing and this was where (the timing industry) was going, we went along with it. Obviously, in retrospect, it was a mistake.

"It was my decision, not anyone else's," Barahal said.

The new system, in which small tags equipped with a tiny transponder replace the larger plastic tags, had been beta-tested at 40 marathons and other races, and was successfully used at the Philadelphia and Las Vegas marathons.

SHORT-CIRCUITS

According to Simms, the problems started hours before the race got under way when heavy showers short-circuited four out of eight generators needed to power the system. That caused each of the remote electronic readers used to track runner progress to shut down.

When the readers, which were synched to the time of day, rebooted, about 40 percent of them "came up with their own time," Simms said.

Simms and his crew manually reset the readers and worked to adjust reporting discrepancies, but problems persisted, causing errant reporting of split times along the route.

"Weather and electronics don't go well together anyway, and we didn't have our equipment weather-proofed as well as we should have," Simms said. "We found that out."

The problems became evident almost immediately as runners began crossing the finish line. Personalized e-mail reports contained wrong or incomplete information. Finisher lists posted on-site were incomplete. Information on the official marathon Web site was also off.

"It was very disappointing," said Martin, whose split times were incorrect and who was unable to find fellow finishers on the results list.

Simms said the decision was made not to include results that could not be confirmed or corrected in the initial reports.

"We had already embarrassed the race enough," Simms said. "We weren't about to make it worse by putting bad data out there."

'IT'S A LITTLE DISTURBING'

Race officials still aren't quite sure how many runners were affected. The official runner counts offer a few red flags.

The confirmed tally of people who registered for the race is 27,829. According to the timing system, about 20,600 runners showed up for the start — a figure that marathon officials already know is inaccurate because there were at least 20,692 finishers.

According to Barahal, between 15 percent and 20 percent of registered runners do not make it to the start line. It is unlikely, Barahal said, that more than 7,000 people elected not to run.

"Historically, we lose 3,000 to 4,000 (between registration and the start) — that's automatic," Barahal said. "The upper end we're working off of, then, is 3,500. That's the credibility gap."

Simms said the finishers' data is probably 98 percent correct; 99.7 is "normal." The data for individual splits is around 95 percent correct, Simms said.

"Five percent is a lot of people," Simms said. "We're pretty good from a statistics standpoint. But from a marathon standpoint, we suck."

Hundreds of finishers visited Kapi'olani Park yesterday to pick up their official finisher's certificates. While most were satisfied that their results were accurate, dozens of others arrived only to find that their times were mis-reported or missing altogether.

"It's a little disturbing," DeWitt said. "At the same time I know I did it, so I have that self-confidence. I don't need a certificate to say that I did it. But it would have been nice to have one."

The Honolulu Marathon Association is collecting reports from runners who think their times were not accurately recorded. Barahal encouraged those with concerns to e-mail info@honolulumarathon.org with their name, race number and estimated finish time.

As of yesterday evening, Simms and his staff were still working to resolve as many of the timing errors as possible. Those that cannot be resolved will be double-checked against video of the race taken for Sports Media Productions' personalized DVD service, MyMarathonDVD.

"This is primarily a people's race and most of our runners are concerned about their finish times," Barahal said. "The split times are nice, they're the bells and whistles, but we focused on getting accurate finish times."

Despite the errors, many who stopped by the marathon tents at Kapi'olani Park yesterday said they appreciated Barahal's candor and the association's attempts to correct the situation.

"This doesn't stop me from anything," said Shin, the runner who saw 6:11 something at the finish line. "They did a wonderful job of managing the course. I have no complaints."

Reach Michael Tsai at mtsai@honoluluadvertiser.com.