The data on which this study is based were subject to right censoring, which means that the published estimates are too low. This was also pointed out in a letter to the editor to the same publication (compare http://www.ncbi.nlm.nih.gov/pubmed/21967106)

Briefly, imagine a litter of 10 puppies born 7 years ago, 3 of whom are dead. With the method used, you only get the dead puppies in your data, who obviously aren't any older than 7, and don't get the 7 puppies who lived to be over 7, which leads to a life span estimate that is too low. This is obvious when you sort the data by birth year, but becomes masked when you sort them by year of death, as the authors did.