Okay, I'll stick my neck out here. My MBA was marketing with a minor in statistics - lots and lots and lots of statistics! It's been a few years - so go easy on me. If the question is how small a sample size can be to accurately represent a population, the answer is 'pretty damn small' - surprising small for that matter.
More important than size (in this case size doesn't matter much) is WHO the sample size is. There is random sampling (whoever is walking by), quote sampling (try and get a representative population) and variations. Who and where the meters are is much more significant than how many there are, in fact, having too many out there can actually DECREASE the accuracy of the reporting.
We'd get into things like standard deviation, chi square values, means, modes and medians that would start to make my head explode and you all go to sleep but we can go there is you want.
Suffice to say the 'problem' if there is one, is likely not HOW MANY, it WHO and WHERE.