Right you are!

The bmi does exactly what it was designed to do.  Every one wants to tell
us how arnnie s. proves it is wrong, or some nfl jock; and I will throw in
fashion models to make the same point in the oppisite direction.  Both are
not normal, they are distortions people express in the normal distribution
curve that bmi uses.  Nor is the question ignored by the "stupid"
scientists who don't know what direction is up, because evryone wants to
see themselves as another arnnie or model who proves that the abnormal is
the rule. In the now outdated charts using "small, medium, large" bones;
everyone saw themselves as "large" who could carry more weight.  The best
use of the bmi is in conjunction with waiste measure, the latter being
more finely tuned to the more specific question of body fat percent and
distribution then the bmi does as a statistical expression that covers
just about everyone just fine.  Consider:

http://www.nhlbi.nih.gov/guidelines/obesity/e_txtbk/txgd/4142.htm

The numbers cover both genders the same, there has been some adjustment
based on populations in some areas where the risk factors upon which the
bmi cutoff,ie. 25 for upper limit for normal, were originally formed. In
those populations it is usually found that the cutoff is best put at 23 or
so to account for the same risk level.

I found it in a piece of software on
http://www.cemax.it/download.html (in Italian)

I've just emailed the author and asked him if he can give me some
reference.

Btw, I tested the formula in a real case and it gave the same result as Dr
Sears' fat mass calculator (in that case at least).

http://www.drsears.com/drsearspages/bodyfatcalmale.jsp