Thanks for more suggestions....

We get the effect of added C directly after blanks (empty tins) and N
standards alone (no C).  Thus, we do not believe it is derived from the
previous sample.  The effect is quite dramatic, for example, 3 consecutive
samples of N11 (28 peaks from 1 - 3 volts) will give us an average of 3.5
plus or minus 0.1 permil, but the following N11 sample spiked with cellulose
(44 peak of 4 volts) returns 7.5 permil.  The same also has been observed
for spiked samples following blanks.

The elemental analysis is fine (i.e., standard curves and QCs are all
great).  After yesterdays suggestions, we tried a new batch of chromium
oxide, but the results were same.  I suppose the only thing left is to tweak
O2 delivery, something we've never had to do.  Is there any rationale or
theory behind the appropriate timing offset?

Thanks for all the input,

Patrick