Unfortunately, the authors of the paper did not understand the relationship between wallets and addresses very well either.

The recommend use of Bitcoin is that every payment you receive should be to a new address. This maximizes privacy and its necessary when you may have multiple concurrent payment in order to sort out which one you've received. Common wallet software (including the reference client) also always pays change to a new address.

... so it's a bummer that the authors of the paper went on to describe every coin assigned to an address which has never spent as in savings (roughly 60% of them), if not for a couple things engaging in the bad practice of address reuse they would have found 100% in "savings".