cuffdiff - replicate samples

[adrian]

Hi:

I have 3 replicates and 4 different samples.

Sample type 1: a1,a2,a3
Sample type 2: b1,b2,b3
Sample type 3: c1,c2,c3
Sample type 4: d1,d2,d3

In Cuffdiff, I supplied all samples and replicates as a list:
a1.bam, a2.bam,a3.bam b1.bam,b2.bam,b3.bam ...

In cuffdiff result, I do not have significant genes.

I want to run two sample type at once . For example comparing a vs. b
a vs. c, a vs. d etc.,

a1.bam, a2.bam,a3.bam b1.bam,b2.bam,b3.bam only.

Will there be any difference in the result. For instance will I loose any power by not giving all samples at once. Is cuffdiff doing some kind of approximation over all samples and then fitting FPKM over two samples.

I appreciate your expert input.

Thanks.