I'm moving your email to the galaxy-user mailing list because it concerns galaxy usage; also, there's a substantial community of users doing RNA-seq that may be able to offer suggestions to help you out.

I used Galaxy (Bowtie) to successfully map 15 million Illumina reads trimmed to 65bp. When I applied Cufflinks to the BAM data no transcripts were reported even though it ran OK. It is possible that very little mapped to known exons in this data set. Does Cufflinks only report data for known transcripts? I thought it was designed to work without a reference annotation. How does it decide what qualifies as a transcript? Any ideas why I got such a result?

Your problems are likely at least partially due to using Bowtie instead of Tophat. Tophat is the standard way to map reads so that Cufflinks can assemble transcripts. Cufflinks assembles transcript--de novo or reference-guided--based on the mapped reads, but mapped reads must include spliced reads, which are generated by Tophat but not Bowtie.

Good luck,