bmi is an epidemiologic tool. it wasn't ever supposed to be used as a diagnosis.
when it is used for measuring body mass in large populations, it works, kinda like a regression line. it's used for categorizing people in large population studies.
however, it overestimates fat in muscular people, black people
underestimates fat amongst asians.
and is best correlated against white caucasians.
doesn't take into account location of the fat (there is a big difference of having an andriod or gynoid shape)
these are many more weaknesses that everyone knows about, but as a tool to measure body mass quickly for large population studies, it is good and correlates well to an r line, not to mention it is very cheap.
there are large variances in individuals, so if you are muscular in real life but come out obese at +30bmi, you are ok and should not worry about what the chart says.
but if you are a complete fatass, and come out 30+bmi, the chart is probably right and complaining that its wrong isn't gonna convince anyone that you aren't really fat.
everyone should inherently know if they are fat or not (anorexics excluded).