After power failure, clock will be 0..10 minutes behind the schedule. If the value written to EEPROM was ahead of actual wall clock by half of timespan between writes (instead of current time), each restart would contribute -5..5 minutes of drift, averaging out to zero over long time and multiple power failures.<p>One could in fact set it to update EEPROM only once a day, but write time half day ahead of current time. This will prevent wearing out memory for hundred of years while still maintaining decent accuracy.