One thing you should be considering is trying to find motifs within your sequences through SEVERAL programs and services. MEME is uncapable of finding all at least in my previous experience.
Try as well using MOODS and many other that will appear after a google search
MEME-ChIP was designed to discover motifs in large number of sequences. I used it to discover motifs in thousands of "short" sequences (peak summits -/+ 200 bp) and it worked well. The limiting factor in your case is the sequences length. With sequences of 5kb to 15 kb of length, you almost discover all motifs and the results will be meaningless. Are you sure you need to discover motifs in such large sequences?