Ok, I'm starting to be convinced that maybe the issues of people doing a lot of heavy training are a problem for the BMI assessment (I still think that'll only be a few % - this forum is not a fair sample of the population). I would hope that for something like joining the TA etc they would use a more complete way of measuring an individual's health/fitness. But I still think it's pretty useful for the majority of the population who do very little exercise. Importantly, if it is an issue around people doing heavy training then I'd have thought the entry requirements to the TA would be a bit more sensible. If it's for whether to refer someone to a dietician checking if their BMI>40 (morbidly obese) might be a useful cut-off. But clearly no one should use a metric like this without taking other things into account, for an individual (e.g. what other morbidity is the person suffering from? what underlying cause is there to the weight-gain?).
The proportion of people in Scotland with a BMI>30 has gone from 17.2% in 1995 to 27.4% in 2010. I really doubt that that is due to 10% of the population taking up serious muscle building regimes[1].
Critcally, if you really think that BMI has no strong association with health, why is it that it is hugely correlated with a massive variety of diseases[2]? It seems a little silly to throw out the whole metric because it's not perfect - no measurement of a population is ideal (either it will be too expensive/impractical/inaccurate/etc..). Obviously subjectively you can say 'woah, that person's big' or 'shesh, you're skinny', but it's very useful to have a quantitative measure of that.
Summary: I agree it doesn't work for everyone, but why dismiss it so completely? People seem to be applying it badly. Which is a fair enough criticism. Also being in the 'overweight' is barely of any concern - concern about being 1.7 over the 'threshold' is missing the point. The thresholds are basically useful tools for researchers/clinicians but shouldn't be abused (e.g. maybe like the TA did).
[1] Scottish Health Survey 2010.
[2] http://en.wikipedia.org/wiki/Obesity-associated_morbidity
(that is a pretty depressing page...).