According to the article you link, the average tenure is one year for users of a site called PayScale. I've never even heard of it, so they have no idea how long I've worked at Google.
He's pointing out that the data comes from PayScale. Most Googlers have never even heard of PayScale, and so the information that is being reported in this article comes from a tiny slice of Google. It's also a biased sample, as someone who goes on PayScale is likely someone who is going to hunt around for the best salary available.
It's dangerous to assume everything you read is correct. You can usually get a hint at how it's biased by reading the article closely.
A number that is skewed by the fast rate of hiring by top-tier companies. When companies are hiring at the rate of Google, Facebook, Apple, etc then even an infinitesimal departure rate of old-timers would set the median at that level.
