I checked the noise floor of multiple recordings I took in the same day in the same room with the same set up. I recorded 10 different tracks all peaking around -6. 4/10 of these tracks had a noise floor of -85 while the other 6 all averaged around -59. I can't think of any reason for this other than the stopping and starting of the PC's fan (no I can't get this out of the room).
Also taking different readings of the same track yielded different NF levels. If I selected the whole track (with noise floor at the beginning) the NF level would be sometimes significantly lower than when I was just measuring the vocal. Maybe the fan was kicking in when I began to talk? Sometimes it would be a small difference. If I am having an issue with varying NF levels (most likely because of the fan), I should only be measuring the vocal, or the parts I will ultimately use, correct?
I use a c1000s AKG, and Focusrite Scarlett 6i6 USB Audio Interface.
Have you done a spectral analysis of the two (high and low) noise recordings? Could you post a sample of both recordings (e.g. 20 sec of each) using the "upload a file" button at the bottom of the reply box? Scale them up first by the same amount (maybe 30dB) to lift them well above the compression noise of the upload process.
What was coming in?
A low impedance mic will yield a lower noise floor than a guitar...
Were you measuring the same thing? As in each input with the gain on exactly the same setting? With each mic in exactly the same place, the same distance and the same brand?