Process analysis and statistical modeling

I am using a XRF device to measure elements in paper sample. The measurement process provides a PPM value and range, in which I can select 1-3 standard deviations. The test can be as short as 30 seconds, as long as 180 seconds. The longer the test, the narrower the error range. My goal is to get as narrow an error range as possible. I need to know if I should be conducting multiple short tests and calculate the range independently or whether to conduct longer tests and average the results.