At the extremes of body fat percentage the estimation methods (which everything is until you die and a mechanical or chemical body fat estimation is done) are not valid for the most part.
I believe the newer DEXA software (and probably MRI and CAT estimation methods) is probably much closer as it has been validated with (and the software is derived from measurements of) animal carcasses and large, low fat animal pieces (limbs) which are then chemically digested to determine fat percentage (about as direct a method as you can get).
Years ago ('98, I think) I stayed in shape after a show to get some BF% measurements done and I was either 4 or 5% on hydrostatic, skinfolds, double-labelled water, BUT
The Bod Pod equations (which have since been corrected) had me at 12%
The torso measurement on a Lunar DEXA machine put my torso (a 2 D scan at that time) at I think -0.4% or something like that. (Yes, a NEGATIVE BF% in my torso... LOL)
So, obviously I did not have a adipose black hole in my trunk or anti-matter in my torso (or whatever the hell it would mean to have NEGATIVE percentage body fat) - it was a estimation inaccuracy b/c the software's algorithms had not been derived from a sample including contest level body fat individuals.
(There more boring scientific dribble behind all of this, but I'll leave it be for now...)
So, the "world record" for lowest body fat percentage would probably be a number (like a negative one) that was derived because of a methodological error moreso than because the person who was measured actually had the lowest body fat (or near it) of "all time."
-S