Some reasons why: The military does not necessarily count the murder of an Iraqi as an attack. If the source of a sectarian attack is not determined, it is not put into the database. And roadside bombs, rocket and mortar attacks that don't injure U.S. personnel don't count. "Good policy is difficult to make when information is systematically collected in a way that minimizes its discrepancy with policy goals," notes the report.
E&P sees this as evidence that the press does not "overstate the level of violence in Iraq," as some critics have claimed. I'm not sure one can directly conclude that from the report's findings, which here, at least, are focused on the military's reporting, not that of the press. What one can conclude is that the military seems to substantially understate the level of violence in Iraq, which is plenty jarring on its own. As for press coverage, E&P is at least partially right – since reporters do rely to some degree on the military's figures, coverage will sometimes reflect the military's underreporting bias. Does this mean that the press never misstates the level of violence in Iraq? No. But it does mean that those who see coverage out of Iraq as unfailingly pessimistic might want to reassess their position.