Battery power estimation and optimization

Hi!

I'd like to hide small sensors in wall clocks and I'm looking for a way to power them. Maybe some of the electo-guys could give me some tipps.

The MCU is an ESP8266 which runs from 1.7 to 3.6V. Most of the time it's sleeping at 1µA, lets say every hour it draws 200mA for 5seconds to do stuff and transmit the data.
The sensors include a DHT22 which needs 3.3 to 5V.

So I need a 3.3 to 3.5V power source. The wall clocks already has one AA battery (1.1 - 1.5V). The best thing would be if I could work with that, so I googled and found step up converters.

Assuming a battery has 0,6Wh (400mAh * 1.5V), the step up thing (maybe this) has an efficiency of 80%, which gives me 0,48Wh and the circuit needs (Sleeping: 1µA * 3.4V * 1h) 4µWh + (5 seconds FullPower: 200mA * 3.4V * 5s) 2mWh = 2mWh.
=> 480mWh / 2mWh = 240h = 10days of operation.

Okay, this is a worst case scenario, but I'd like to do the math with realistic values. To get better performance, I could use better batteries with 1200mAh, which would give me 30days of operation. If I use two "better" batteries in series, that probably would increase the runtime to 2 months (assuming the clock won't draw that much current) which would be okay for me.

1. Are those assumptions and estimations roughly correct?
2. Are there better ways than a step up converter? Maybe put three batteries in parallel and use a low dropout, low quiescent voltage regulator (maybe this). Since the efficiency of the regulator would also be 80%, is the result the same?
3. Is it correct that it's not a good idea to put three batteries in parallel and grab 1,5V off the first battery to power the wall clock? This means, if I decide to use the "batteries in parallel" approach, I need a separate battery for the clock?
4. In general: How would you do it?

I considered using a rechargable LiPo but I think it's weird to recharge a dozen wall clocks every few month :). Also I don't like the protection circuit.