Ok, I missed parts at the bottom of the first post. Does your fasta file have transcript sequences? I'm assuming then you have transcript sequences that contain intronic regions that aren't spliced out? Or are these simply splicing variants and alternate exons? Either way, you can use regular expressions to grab the exon lengths in the cufflinks file, and you can compare this to the sequence length from the fasta file or other transcipt information files from the UCSC genome browser (or the GTF file that you are likely using from Illumina via the cufflinks webpage to annotate transcripts), etc. Again, you can store the transcript information in a hash and use the gene symbol or accession number as the key so that you are comparing the correct things.
Sorry for missing that, I need more coffee it seems :).