My question is, if I use the public Galaxy server interface to TopHat and Cufflinks, is there any access to "cuffmerge"?

Also, I'm trying to understand the difference between using cuffmerge and then using cuffcompare (without a reference genome) to assemble gtf transcript files produced by Cufflinks for each group of 3 Illumina paired-end reads corresponding to biological replicates, in order to use the resulting combined gtf file for comparing the TopHat alignments of two such groups using cuffdiff.  

Is there any difference in the output between cuffdiff and cuffcompare, using in this fashion?  For example, do they form the "union" of transcripts by the same rules, and do their outputs contain (or lack) the same columns (strand, perhaps??)  I've read things on seq-answers indicating that I should be using cuffmerge, but I can't find it on the public server and apparently haven't installed it properly on my own computer so far.