Detailed information is required.
I don't use a PE instrument (LECO Truspec and TruMac), but I'm guessing that the setup is similar.
Are you getting high or low values ?
Does the high/low value change over time ?
Is the N and S there in the blanks or an empty combustion ?
What calibration standard do you use and is the sample values within the cal curve ?
How do the values of certified references hold up ?
What warm up time do you use/do you turn the instrument on and off ?
Some aspects to look at could be gas contamination, calibration curve linearity (and intercept at low levels) and calibration standard contamination. You should be using Ultra High Purity gases, usually stainless steel pipes for the carrier and copper for the oxygen (although we seem to be having problems with our oxygen and I'd had one indication that it could be nitrogen diffusing through the copper pipe, but I've not heard of that before). If possible, you could try turning the oxygen supply off at the cylinder or in the software and then running a few blanks to see what happens.
Have a look at the calibration curves and see how they appear at low levels. Are the peak areas within your standard range ? Can you change the cal curve type from linear to polynomial/cubic/quadratic or single point and if so, do the values change significantly ?
As I said, I don't use a PE instrument, so some of the above may not be relevant.
Shane.