There are two issues w/ regards to the mountain bike/triple ring version of the SRM:
1) the distance from the sensor to the inner face of the powermeter (crank) is greater, meaning that you have to increase the strength of the magnet in (of) the sensor so that the powermeter will "wake up".
2) more importantly, the inner chainring is significantly off-center with respect to the plane of the powermeter, i.e., when looking down the granny ring is offset towards the centerline of the bike. This can causes problems with calibration/power calculation, since the sensitivity of the strain gages to such off-axis torques may be different as a result. Unfortunately, I cannot tell you the magnitude of the possible error, only that I've heard about it. (I can say that my SRM Pro road crank shows a systematic ~2% variation in slope factor, depending on the chainline.)
If maximum accuracy is needed, you might be able to back-calculate the rider's gearing from their speed and cadence, and then apply a chainring-specific correction factor to the data as reported by the SRM.