Republican John W. Warner's apparent victory over Democrat Andrew P. Miller is unlikely to be erased by official canvasses of the state's vote, a member of the election news agency that gathered the unofficial - and controversial - vote totals said yesterday.

Jean Cryor, mid-Atlantic manager for the consortium of news agencies and networks that processed the election-night results, acknowledged that computer breakdowns Tuesday night and subsequent confusion cast doubt on the accuracy of Warner's final unofficial 4,512-vote margin look fragile indeed.

But survry of the vote begun by election officials yesterday reported minor, if any changes between the previously reported results and official tallies. Local offcials in Northern Virginia and elsewhere said that Warner was gaining a few votes and widening his lead over Miller.

The sharpest change reported yesterday came in Albemarle County, outside Charlottesville, where Miller lost 639 votes in an official canvass and Warner gained 30 votes.

In Fairfax County, the state's most populous jurisdiction, a canvass completed last night added only two votes to both Miller and Warner's totals, an elections board official said.

Cryor, an official of the News Election Service said however, the unreliability of the previously reported Virginian results is more apparent than real. She said the accuracy of NES results had been double-checked Tuesday night and Wednesday.

"I stand behind those figures," said Cryor, an eight-year veteran of the vote-tabulating process. "Maybe I shouldn't but I do."

Separately, Warner campaign spokesman William Kling said the Warner officials were growing more confident that the former Navy secretary's victory would stand.

Warner himself was secluded at his Fauquier County estate yesterday. He has refused to claim victory and Miller, a former state attorney general, has refused to concede the elections until the outcome of a Nov. 27 canvass of the election.

Final NES figures released Wednesday afternoon showed Warner with 613,294 to 608,782 for Democrat Andrew P. Miller - a margin of about one-third of 1 percent in the closest general election in the recent Virginia history.

Besieged by calls about possible inaccuracies in the NES figures and questions about a possible recount, state officials Wednesday called a news conference to point out that the state officially has nothing to do with the unofficial NES returns.

Officially, no winner can be proclaimed until Nov. 27, when the Virginia State Board of Elections makes public the results of its official canvass under a process governed by state law.

Virginia Gov. John N. Dalton noted Wednesday that unofficial returns in his race for governor last year had shown him to have some 7,000 votes more than the official canvass later proved him to have.

Cryor pointed out yesterday that NES did not tabulate those figures. The service, a network of thousands of people nationwide, is set up only for federal elections where, she said, its accuracy has been proven during its 10-year history.

Initially, uncertainty over the Virginia vote centered on two factors: NES was listing more precincts than the 1,857 acknowledged by the State Board of Elections, and the service produced different sets of running totals during election night computer trouble.

Cryor said neither factor affected the accuracy of the final returns. The precinct discrepancy, she said, was caused by an arbitrary NES procedure that groups absentee ballots in some localities as separate precincts for the purposes of tabulation.

The discrepancies in the running totals, she said, were caused by different tabulating procedures being used by the main NES system in New York and its backup system in Richmond.

"We didn't just have one period of computer trouble," she said. "The (computer) system in New York kept going up and down all night."

Whenever it went down, she said, NES would switch to the backup system where statewide totals were being compiled on a slower county-by-county basis instead of precinct by precinct as in New York.

What appeared to be differences in vote totals, she said, were actually differences in tabulating speed between the two systems. Ultimately, she said, the same votes were counted by both systems until about 2 a.m. Wednesday when the Richmond center was shut down and the tabulation was switched to a New York system now functioning normally.

"But just to be sure, I personally called every county or city official (in charge of the vote) independent of the computer and added up their totals on an old-fashioned adding machine with a tape," Cryor said. "I'm sure there's probably a vote or two out there somewhere but nothing substantial.

"In a state with as many counties and citied as Virginia (140) in order to overturn a lead the size of Warner's you would have to have a widespread pattern of error all tilting in one direction."

Yesterday's vote canvassing was a tedious process particularly in those areas of the state with larger votes turn-outs. Some jurisdictions continued to check over precinct-by-precinct returns well into the evening, while others recessed their activities late yesterday and were expected to complete work by today.

Only one election district in North Virginia, Arlington County, reported any change that differed from their unofficial election night tabulations in the Senate race. In that instance, Warner picked up 28 votes to register a new Arlington County total of 18,938 while Miller's total remained unchanged at 23,772.

In the Richmond and Tidewater areas of the state with larger voter small vote gains in election districts there but Warner appeared to offset these with new increases of his own.

In Isle of Wight and Portsmouth, for example, Miller gained 20 and 212 votes, respectively while Warner's tally stayed the same in Isle of Wight and only increased by one vote in Portsmouth.

In Richmond, Miller's showing declined by 247 votes to 31,503 but Warner's vote, total dropped by 250 votes down to 27,743.

And in Chesterfield County, outside Richmond, Miller and Warner both lost votes in the 3rd Congressional District section - 60 votes less for Miller and 164 votes less for Warner - but each picked up more than 100 votes in the 5th Congressional District section of the country.

Overall, the changes "have been very, very minor: a vote or two, something like that," Kling, the Warner spokesman, said. "We think it's (the margin) going to hold."