I've sometimes used a value of about 1500 W/m^2 as a flag for unreasonable data. This has a little bit of wiggle room from the solar constant, but at least draws your eye to the data that you may need to verify. It's a tough call whether to modify those values in the weather file, or maybe just exclude those timestamps from the analysis completely.
If one of your irradiance components is unreasonable at a given timestamp but the other two seem correct, you can always use the reasonable two to calculate the third using:
GHI = DHI + DNI * cos (Z)
where Z = zenith angle (you can use the zenith angle output from SAM if needed). More info on that equation can be found in the definition from the
NREL RREDC Glossary of Solar Resource Terms
. Hope that helps!