High CpG vs. low CpG

I think it might be good to bring up the figure I just read from the following paper:

DNA Methylation and Structural and Functional Bimodality of Vertebrate Promoters. Navin Elango and Soojin V. Yi. Mol Biol Evol (2008) 25 (8): 1602-1608.doi: 10.1093/molbev/msn110

It's known that the normalized CpG score of promoter regions has a bimodel distribution, with the local minimal at around 0.4. But it's striking that this is distinct for invertebrate genes - the promoters of see squirts have unimodal distribution and its introns have bimodal distribution.