It's in the scientific literature if you want to take a look you could probably google it and you'd find decent references. I guess is what it is really saying is the healthy range is a little different to this 20-25 range we hear bandied about all the time.
We regularly take this measurement in clinic along with waist;hip ratio and some people have BMI's of 23 and you look and think - you're fat (using the actual definition of the word fat (tissue type) but you're not overweight! However, you could still stand to lose a little fat.
Others come in with a BMI of 27 and you at them and are pretty sure they're healthy and don't have any fat to lose - you couldn't pinch an inch anywhere so to speak.
My BMI is just under 25 right now, I'm tall and this is heaviest I've ever been and whilst I don't like busting out of size 12's, you wouldn't look at me and say fat! I would like to lose 2 stone to get back my lithe body shape (starting low carb diet tomorrow), look much better with a BMI of around 22.
My husband was on a study that was being run by my colleague and his BMI put him at almost obese, and yes he has a small middle-age spread type belly but there is no way he needed to lose 3 stone to get into normal range. He'd look terrible. He's healthy as a horse and his metabolic profile proves it.
So YANBU at all to dispute it for yourself and there will always be exceptions to every rule. But on an epidemiological level, BMI is a very useful indicator and as such, the 20-25 range stands as a reasonable one.