cufflinks quantification problem

[Trudy]

Hi all,

I have a question about quantification with Cufflinks.
I've performed a quantification analysis with Cufflinks in two ways.
I've 4 samples for Population (samples are biological replicates) and I did:
-quantification analysis for each sample
-quantification analysis for population (before I performed samtools merge to join 4 alignment files)

How can I connect these results?
The FPKM values are very different. I expected that the FPKM values for population were le mean or something similar of the FPKM values come from each quantification samples.

And also, the FPKM values from differential expression analysis (with cuffdiff) for each population are different compared to FPKM population values....which is the criteria to determine FPKM in differential expression for Cufflinks?

Thanks in advance