I posted a few days ago about a stock DS1307 design I used in a '328P clock display circuit.
The source of the problem was found - my customer was testing with 4.7 ohm I2C pullup resistors, and not 4.7K ohm I had in the parts list. So while my properly equipped test board worked fine, his would drain a coin cell dead in 24 hours.
I don't understand the path for that to happen within the DS1307, but happen it did.