My question comes down to:

At very light loads (1uA@5V), is a good switching regulator still more efficient than a very low IQ linear regulator?

Now, on to the specifics, which I would appreciate you checking my understanding on...

So, I've got a project that operates off of a Ultralife U9VL-FP 9 volt primary lithium (not rechargeable) battery. It has a voltage between 5.4 to 9.9V and capacity of 1.2Ah @ 9mA to 5.4V at ambient.

My system requires 5VDC with either a 60mA or 140mA load. The reason for the disparity is that I haven't decided how much juice I want to give the contrast on my display :-) The device will be used at this load for about 2 minute spread evenly throughout the day (not night). So maybe 24 total minutes per day.

When the system is "off", the microcontroller must remain powered and should be running at <= 1uA. So 99.9% of the product's life is spent at very low current load, but the regulator MUST remain on.

I know that when the device is on (120mA), the switcher is deffinatly going to be far more efficient. But, which is more efficient at light load (1uA@5V)?

Small (package, caps and inductor) is important to me, however cost isn't.

On the linear regulator side, I'm looking at the National Semiconductor LP8345

With an input voltage of 9.9V and a quiescent of 19uA and then 1uA @ 5V output, we know that the regulator is 26.6% efficient.

On the switcher side, I'm looking at the Linear LTC1877

However, the datasheets for the linear switcher are very deceiving, giving NO input quiescent specs at 1uA output. Their Input DC Bias Current spec is misleading, but the best I can gather would be that with VIN=9.9 and VSYNC/MODE=9.9, I would gain a typical quiescent of 10uA. If that's the case, then it would be running closer to 50% efficiency.

Also, are my (part) choices good? (remembering that I don't care about cost, but small size and minimal numbers of small externals I do care about)

Thanks
Tomas