The relative accuracy (RA) of a newly developed mercury continuous emissions monitor, based on X-ray fluorescence, was determined by comparing analysis results at coal-fired plants with two certified reference methods (American Society for Testing and Materials [ASTM] Method D6784-02 and U.S. Environment Protection Agency [EPA] Method 29). During the first determination, the monitor had an RA of 25% compared with ASTM Method D6784-02 (Ontario Hydro Method). However, the Ontario Hydro Method performed poorly, because the mercury concentrations were near the detection limit of the reference method. The mercury in this exhaust stream was primarily elemental. The second test was performed at a U.S. Army boiler against EPA Reference Method 29. Mercury and arsenic were spiked because of expected low mercury concentrations. The monitor had an RA of 16% for arsenic and 17% for mercury, meeting RA requirements of EPA Performance Specification 12a. The results suggest that the sampling stream contained significant percentages of both elemental and oxidized mercury. The monitor was successful at measuring total mercury in particulate and vapor forms.