au contraire
I would like nothing more than this figure to be two or three orders of magnitude too high. But if wishes were horses....
I don't think the survey differentiated between completely innocent civilians and insurgents - family members reporting deaths of loved ones don't likely see a lot of difference, and in fact the surveyors purposely didn't ask them to differentiate.
Separation of combatant from non-combatant deaths during interviews was not attempted, since such information would probably be concealed by household informants, and to ask about this could put interviewers at risk.
(from the Lancet study, p. 2)
My example of two or three Iraqi casualties per attack on US/coalition forces was obviously high as it didn't allow for large numbers of casualties in instances of intensive fighting.