I'm not sure if you mentioned this - I skimmed a bit of the OP, sorry :0 - but is there a figure or profile for the sensitivity of the test?

Do samples really matter that much here? Unless I'm missing something (a definite possibility!) what it seems you want to do is ascertain whether a specific razor contains a certain element or not, and in what proportion if the former. Surely that simply requires a "presence" cutoff higher than the measurement error?

If you are trying to infer something about the population of all "silver steel" razors ever produced, that's when sample size (and other things like representativeness of the sample) would definitely be important.

Again, apologies if I'm not really understanding your research question here.

James.