A lot of people might have created Bitcoin wallets for fun (it takes just a few seconds). But these wallets are all empty, with no transaction ever occurring with them. I am afraid that the survey might have counted these wallets as well.
Most wallets do that automatically for convenience purposes. I however don't for a second think that they counted empty addresses because it's meaningless on its own as anyone can generate addresses out of thin air at any time.
there is no possible method for counting unused addresses anyway.
unless the address is used, it doesn't appear in the blockchain. the only way unused wallets could be miscounted like that would be if a webwallet did so, but you'd be unwise to believe any such unverifiable claims made by webwallet sites, they're pretty heavy on the "marketing rhetoric" (i.e. bs), as you would expect