guys, IMHO you are chasing diminishing returns powering a DS18B20 from a DIO pin.
Data sheet says DS18B20 stand-by current is 1μA max, 750nA typical. The leakage current on a ATmega328P DIO pin can be up to 1μA, and the chip itself draws at least 4.5μA in sleep mode.
I have several temperature measuring nodes on my JeeNode network (
http://jeelabs.net/projects/cafe/wiki/ - JeeNode is similar to Moteino). Once per minute they each sample temperature and their own battery voltage, and send the data to the monitoring station via the RFM12B. They are powered by a single AA alkaline cell via a 3.3V booster (
http://jeelabs.net/projects/hardware/wiki/AA_Power_board), so current draw from the AA is about 2.5x the current that the DS18B20, μC and radio chip draw directly from 3.3V. The DS18B20 is connected directly to the 3.3V rail and uses a 4k7 pull-up on the data line connected to the DIO pin. In other words, no special power conservation measures.
One of these nodes was installed with a partly spent AA that was reading 1.2V when put into service. After running continuously for 15 months it is
still going (cell voltage is now down to 0.9V). The nodes that started 15 months ago with a full AA cell are now reading about 1.3V.
So even with 4 DS18B20s attached, you should get >1year battery life as long as the μC is programmed carefully. For example, be sure to put the μC back to sleep while it waits the 750ms needed for the DS18B20s to do their 12-bit conversion, and make sure the DS18B20s all convert at the same time to minimise the number of times the μC has to wake-up.