That chart doesn't show retention of new users. It shows active users. For example, if you get one new user per day, each is active for a day and then quits, but one of those users sticks around past the week, a daily-active-user chart similar to the one above would be a flat line at one over the course of the week (or perhaps briefly bumping up to two) but retention is actually 14%.

It is very hard to measure because we don't know how many new accounts created (and did not persist as active users) ever corresponded to users, but we do know that many did not.