Methylation of DNA tends to persist through cell division because there are enzymes which approximately copy over the methylation patterns during DNA replication. If some epigenetic changes happen in sperm or eggs, then some of this may be carried over to offspring. You're right that we're still in the early days of mapping this out and understanding it. Molecular biology is a crazy mess that makes the worst spaghetti code look downright reasonable. That's evolution for you....
This is an oversimplification, but imagine you have a sweater that you dye bright green. You then unravel it and use part of the yarn, along with yarn from another sweater, to knit a new sweater. Some of the new sweater will have the same physical change, the bright green dye, that the old sweater had.
Sure, but only cell#1 has the actual molecules from the parents.
Cell division requires that the DNA be replicated. What surprises me is that the replication step (apparently) preserves methylation. (If it didn't, there would be no observable epigenetic effect in the organism).
i.e. from an information perspective, the bits encoded in the DNA aren't don't just come from the sequence of bases, but also "out of band" data.
One rather relevant example from a recent new scientist that, in rats, overweight dad rats had offspring who more likely to get diabetes.
I get the impression it's early days with all this stuff, but it's going to be interesting.
http://en.wikipedia.org/wiki/Epigenetics
http://www.sciencedaily.com/releases/2005/10/051026090636.ht...
http://ndt.oxfordjournals.org/content/24/4/1088.full