Take a look at the signal at the ouptut of the AD8362 on an oscilloscope. I think that you may see some ripple. Your carrier spacing is 600 Hz (i.e. one carrier is 2.001 MHz and the other carrier is at 2.016). That is a very slowly changing envelope. My guess is that you are not doing enough averaging to get a valid rms computation. So my suggestion would be in increase the cap on pin 9 (CLPF) until the output envelope goes away and you have a dc signal.