Back in 2009, I had the honor of working with Dr. Keith Prufer of the University of New Mexico and Adam Nazaroff of Stanford University to look at the reliability and validity of fancy new portable x-ray fluorescence devices that were then all the rate in archaeology. We did some fancy statistics to compare the handheld devices to there larger laboratory cousins, our results are below.
To translate: it worked pretty well! The only problem was that the handheld produced values that were a little offset from the laboratory machines. Fixable - but you still *should* send a few samples off to the laboratory folk to get a handle on error that may be introduced.