When a cell divides, the methyl groups are only on the original strand; the new complimentary strand doesn't have any. The methylation signal has to be actively transcribed from one strand to another; an enzyme runs up the DNA feeling for methylated cytosine residues. When it finds some, it starts methylating any cytosine residues that might be nearby on the opposite strand, to make sure the troublesome regions all stay commented out. That's why it's heritable.
The methylation inactivation is heritable. The issue, in this case, was erroneous activation or switching of cells to modify protein production.
I suspect that the mechanism involved (they don't say) in the repair of the genes which end up going back to normal is related to the production of O6-methyl-transferase via the MGMT complex sites on the long arm of c21 -- the same thing that results in chemo-resistance to cancers, such as pancreatic cancer or glioblastoma, when combined with the appropriate mutation of the p53 gene on c17.
I think as long as it doesn't involve a long term mutation of a cancer related gene, such that it effect the germ cells, it's not a problem. Since you tend to come pre-packed with all the germ cells you are ever going to have in your lifetime, then the issue will be smoking by pregnant women, and all other damage that results in disease will only be self-inflicted diseases, rather than heritable.
Which still means they've failed to answer the question of whether or not it's heritable, because they've failed to discuss whether or not it impacts germ cells (arguably unlikely, but it'd be nice to have an answer, particularly when making decisions on how and when to regulate smoking, or minimally, smoking in public).