Drones - the impossibility of accurate statistics

Whilst accepting that US drone strikes in Pakistan's Federally Administered Tribal Areas (FATA) - and elsewhere - regularly result in the deaths of non-combatants as 'collateral damage', I have always argued it is impossible to be precise about the numbers.
Reports in the last year from the Bureau for Investigative Journalism, Stanford Law School/NYU School of Law, the New America Foundation and the Long War Journal all rely on figures provided to local reporters by unnamed security officials - usually members of Pakistan's ISI, who refuse to divulge their identity. Outsiders cannot check these figures as they are banned from entering FATA. The US authorities, which claim that civilian deaths are minimal, fail to provide any evidence to back up their argument.
All the various analytical reports attempt to portray themselves as authoritative, although all use different methodologies.
The debate on drone casualties is rapidly becoming a classic debate on the use and abuse of official statistics. The latest contribution comes from the Human Rights Clinic at Columbia Law School, whose report, Counting Drone Strike Deaths has produced a further set of estimates - although it makes a point of stating the impossibility of accuracy: “Drone strike casualty estimates are substituting for hard facts and information about the drone program,” says Naureen Shah, Acting Director of the Human Rights Clinic at Columbia Law School, adding: “These are good faith efforts to count civilian deaths, but it’s the US government that owes the public an  accounting of who is being killed, especially as it continues expanding secret drone operations in new places around the world.”
Columbia Law School also recently published The Civilian Impact of Drones: Unexamined Costs, Unanswered Questions, which examines the way drones are rapidly becoming the centrepiece of US counter-terrorism strategy.

Disparity in official figures for civilian killed by drones

How many civilians in Pakistan have been killed in drone strikes? Leaving aside the question of who is a civilian and who is a combatant, there is an enormous disparity, as pointed out in an article by Justin Elliot of ProPublica - not least, in those given in off-the-record briefings by Obama Adminstration security officials.
Bill Roggio, editor of the Long War Journal website, which bases its estimates on news reports, puts the number of civilian killed in Pakistan at 138, against 2,307 Islamist fighters. The New America Foundation states that, based on press reports, between 293 and 471 civilians have been killed in the 300-plus attacks since 2004. 
The London-based Bureau of Investigative Journalism, which draws on sources including researchers and lawyers in Pakistan, puts the number of civilians killed at between 482 and 832. The authors of the various estimates all emphasize that their counts are imperfect.
Elliot has used another indicator to analyse civilian casualties. He has compared the many claims made by Obama Administration officials about the number of civilian casualties. His research shows that a wide variety of figures have been given out - off the record, of course - over the last few years. These include "a handful", "zero", 60, 50, "about 30", "under 40", "at most a few dozen", "roughly 30" and so on.
US security officials - the CIA - claim they know the figures because they monitor phone and other conversations after a strike and they observe funerals and other events connected to the strikes. Sounds good, but with such a variation in figures, their case is undermined.
For some people even a single civilian death is one too many whilst for others the issue is that of proportionality - whether or not the number of civilians being killed is proportional to the military advantage anticipated. For those running the CIA drone campaign in Pakistan it seems that the only figures that matter are the number of senior al-Qaeda and other militants who are being killed.