I'm designing a mixer, and I'm seeing something interesting in the noise summary output of a pnoise simulation in spectreRF.  If I try to look at the noise summary for the integrated output noise for a set bandwidth (between f1 and f2, say) without any noise sources excluded and no truncation, the resulting noise magnitudes and percentages do not sum up to 100%!  This is true for both noise power as well as noise voltage.  In essence, I get a list of noise contributors that quickly dwindle to zero (even with extremely high precision using the noiseSummary command from ocean).  However, summing the noise sources listed does not match the listed total output noise, and the percentages listed alongside each contributer do not add to 100%.  However, the summary is consistent in that if the listed contributor percentages sum to 55%, say, the sum of the noise voltages listed is 55% of the total output noise voltage listed.

My question is where is the noise that is not accounted for coming from?  It's not clear to me where it is coming from.  Is this a problem other people have seen?  Is it ok to ignore it, or should I be concerned?

Thanks in advance.