[galaxy-user] cuffdiff values different for same sample

Hello forum, I have ran cuffdiff with two samples, treated vs untreated, and had certain values as expressed in the output (isoform differential expression). When I ran it again, using a different treated sample, but SAME UNTREATED SAMPLE, the values assigned to the untreated are different. Why is this if values are calculated from a unique FPKM? This happend for other jobs where I have ran the same untreated vs other treated samples... Thanks a lot, ib