A political officer in the fortress-like U.S. embassy here last week bent over charts and graphs mapping statistics he had meticulously culled from Salvadoran newspapers for the past 15 months.

Two blocks away a Salvadoran too fearful to tell a visitor his name showed long lists of names and other information written in cramped longhand in red ink.

In the throes of El Salvador's 2-year-old rampage of political violence and civil war, the two men are compiling the country's most politically crucial statistic: the body count of noncombatant civilians.

By anyone's count, the numbers are horrifying. The lowest calculation is the U.S. Embassy's: 7,372 people killed since the embassy began to keep track in September 1980.

The highest estimate was reported by Salvadoran Defense Minister Gen. Jose Guillermo Garcia, who told local newspapers two months ago that the violence had claimed 30,000 victims since 1979--24,700 of them "people who had nothing to do with the conflict."

At stake in the reporting of the deaths is more than the numbers themselves. Each side in the political struggle wracking the country has blamed the other for most of the brutality. The documentation of the deaths and their circumstances, along with the methods and sources employed by those compiling the statistics form the basis for answering the question of who bears most of the blame.

Garcia's figures for civilians, which were announced without elaboration, are on the same order of magnitude as the findings of the Catholic Archdiocese of San Salvador's Legal Aid Office--8,062 in 1980 and 13,353 last year--and the Jesuit-run Central American University calculation of 12,500 and 13,500 in those two years.

According to these estimates, at least one of every 215 people has been killed in this country of 4.6 million since the escalation of violence following an October 1979, military coup.

The task of assigning responsibility to one side or the other has been so fraught with political implications that many news reports have resorted simply to citing the estimates and attributing the deaths to undefined "political violence" of the left and the right.

Not so the compilers. The archdiocese's Legal Aid Office, in its "Statistical Balance of the Repression," attributes 60 percent of the deaths to joint actions of the Army and security forces and 35 percent to "unidentified paramilitary gangs" of the right. There is no category for victims of leftist violence--a deficiency that has evoked criticism of the office by San Salvador's Bishop Arturo Rivera y Damas.

The university's Documentation and Information Center concluded that there is reliable evidence pointing to what group was responsible in only 30 to 40 percent of the cases of civilian murders. Of those cases, the center found that four out of five were killed by the government forces and rightist gangs working with the acquiescence of the military.

The confidential U.S. Embassy report, a copy of which was obtained with embassy authorization, is a compilation of the weekly cable to Washington on violence, nicknamed the "grim-gram." The report found that responsibility could be reasonably determined in about 29 percent of cases, with government forces responsible for 46 percent, rightist "terrorists" 7 percent, and leftist "terrorists" 47 percent.

The embassy report, which officials said is based entirely on information published in the San Salvador press, comments that "responsibility for the overwhelming number of deaths that occur here is never legally determined nor usually ever accounted for by reasonably clear and coherent published evidence."

While contending that "El Salvador's tangled web of attack and vengeance, traditional criminal violence and political mayhem" make it impossible to sustain the charges that the government security forces are the "primary agents of murder here," the report adds that there is "no attempt to lighten the responsibility for the deaths of many hundreds and perhaps thousands which can be attributed to the security forces."

An embassy official acknowledged that the study's total number of deaths is a minimum figure and that the actual number could be up to 30 percent higher.

The report found an "encouraging . . . decrease" in violence for the September 1980 to September 1981 period that is the main body of the survey, but the official said that the level of violence jumped sharply again in November and December and has continued upward in January.

The embassy findings are a significant departure from charges frequently made by officials of the Reagan administration over the past year that the overwhelming majority of the civilians have been killed by leftist guerrillas. In congressional hearings last February, John Bushnell, then the acting assistant secretary of state for Latin America, charged that guerrillas were responsible for 6,000 of a total of 10,000 deaths reported up to that time.

Bushnell said his figures were based on "public announcements" of the leftist guerrilla groups. The guerrillas have denied they had made such announcements, and human rights organizations disputed the U.S. figure.

The Salvadoran civilian-military junta also has claimed that the left is responsible for most of the killing, but has not made public any statistics to back up its case. Junta member Jose Antonio Morales Ehrlich said in an interview that his government is new and does not have the "financial means" to organize a human rights program to document the violence.

Those blaming the government forces, he said, were operating from a "naive" and "romantic" concept of human rights or had been fooled by "disinformation of the Soviet Union."

He said that even in cases in which civilians were killed by government forces, guerrillas were ultimately responsible because they had started the killing first and provoked the military reaction.

A human rights worker who asked to remain anonymous said that the evidence pointing to the military and allied paramilitary groups as primarily responsible was "not just circumstantial," but neither could it be totally conclusive. The worker said that gathering information about killings beyond local newspaper accounts was extremely difficult and dangerous because it involved compiling personal testimony from survivors or witnesses, then protecting the sources from reprisals .

Some Legal Aid Office statements on the violence and its causes several months ago were severely criticized as politically biased by acting Archbishop Rivera y Damas because of their failure to report violence by the left and for making strongly political statements against the government.

A high official in the archdiocese said the criticisms were directed "at their political statements, not at the quality of their investigation." He said the church continued to regard the documentation of the body count as "scrupulous," although clearly not impartial in its omission of leftist violence.

Legal Aid workers declined to explain the methodology of their studies except to say that they had obtained personal testimony for every body counted. Other sources said the office has workers stationed in cities all around the country to compile data.

The embassy report, since it is based exclusively on newspaper reports, provides a minimum figure for the killing. The embassy official acknowledged its limitations. The study excludes from its compilations deaths classified as casualties in military action, he said. But the official said that guerrilla attacks are not considered military action but rather "terrorism" and therefore any resulting deaths are counted as civilians killed by the left.

The study's definition of military action is limited to offensives initiated by government forces, and resulting deaths are considered combat casualties. The embassy study does not include in its count any civilians killed as a result of government combat actions.