Fancesco,
Its not a problem, its the whole point of the exercise. Note the title to Gallego's presentation, namely increasing Signal/Noise ratio (SNR) through rejection. My understanding is that since that presentation was done, there have been updates and name changes in the ImageIntegration tool. It is also my understanding that the number he uses for what you are trying to match up between the no rejection case and the rejection is not the SNR estimates but instead the reference noise reduction. That number is the theoretical maximum benefit you can get from noise reduction. If you go above that number you are rejecting signal and if you are below you are not rejecting as much as you can. But the goal throughout is to use the rejection tool to increase SNR which is what the exercise is all about.
Looking at your particular case, you are close since the reference noise reduction numbers are close in both cases, though you could tweak your rejection settings to get even a little closer. My rule of thumb, however, is if you apply more rejection (i.e., use a lower rejection number) and you see no improvement, then use the highest setting (with the least rejection) that gives the same result to avoid eliminating signal.
Hope that helps,
Best,
Jim
