Some of the TFs (e.g. Bat and some of the tbp's) are probably bad IPs and will ultimately be removed.
The network can (and should) be filtered to get any meaning out of it... For example, the top-level network has edges with a P-value of < 0.2, but the subnetwork was filtered to only include edges with p < 0.05.
You might find it useful to further filter it using other edge attributes (and Cytoscape 2.6's nice filtering capabilities):
"Distance" (how far the hit was from the start-site of the gene)
"where.hit" (e.g. to only include "upstream" hits vs. downstream or in-gene) -- this is also used to set the edges' colors.
intensity (intensity ratio of IP vs. WCE of the peak)
Genome Browser (some data available to Halo group only for now)