reading ATMEGA's low sleep current

Hello,
I have a bread-board prototype of a wireless battery powered Arduino sensor. I used a Radioshack 22-811 multimeter to read the current consumption. It's not the best multimeter out there, but I thought it'd be good enough to read it. I got about .1mA, or 100uA sometimes. But sometimes the multimeter seems to interfere with the ATMEGA328, and the microcontroller doesn't respond correctly, and I see readings in the several mA because it's not going to sleep.

So this made me doubt the 100uA reading I was getting when it was responding correctly. The uC sleeps, and wakes up on an interrupt, sends a wireless packet, then goes back to sleep.

I've read about Dave Jones' uCurrent reader, but I didn't think the Arduino sleeping would have low enough current consumptions to need it. The manual for the Radioshack meter doesn't really tell you the burden voltage or anything I found useful for determining what current levels should practically be able to measure.

http://www.radioshack.com/graphics/uc/rsk/Support/ProductManuals/2200811_PM_EN.pdf

Reading about sleep currents, I'm also unsure if my 100uA is reasonable. I'm doing SLEEP_MODE_POWER_DOWN, and sleeping both the wireless transceiver (RFM69HW) and the ATMEGA. The RFM69HW lists sleep current at 10uA. I'm using 4xAA batteries (about 6V output), feeding a 3.3V MCP1702 regulator. So the whole thing runs off of 3.3V, 8MHz oscillator. Other than the wireless transceiver, I have a voltage divider for reading the battery voltage (2x2Mohm resistors w/ .1uF cap on the lower resistor).

Does my 100uA reading seem reasonable?

On a low current range the multimeter will exhibit a lot of resistance, which will
indeed interfere with normal operation.

Add a switch across the multimeter terminals to short it out.

Once the micro is in sleep mode open the switch to measure the current - at this
point the low current draw means the high resistance of the meter isn't an issue
as the voltage drop across it must be small.

Use two multimeters. One in series with the load (Chip) to measure the current, and one across the chip supply pins to measure the voltage under load (and be able to adjust it accordingly)

// Per.

Zapro:
Use two multimeters. One in series with the load (Chip) to measure the current, and one across the chip supply pins to measure the voltage under load (and be able to adjust it accordingly)

// Per.

The issue is that a multimeter on lowest current range (100uA) has significant
resistance, perhaps 1k or so, so cannot power the device when its awake. Adding
another meter in parallel doesn't address this.

MarkT:
The issue is that a multimeter on lowest current range (100uA) has significant
resistance, perhaps 1k or so, so cannot power the device when its awake. Adding
another meter in parallel doesn't address this.

Yes it does. Usually the input impedance of a multimeter is between 1 and 10 Megaohms, so it will not affect the measurement. Putting it across the load, makes it possible to adjust the supply voltage, to cater for the loss in the shunt resistor found in the multimeter measuring the current - resulting in a proper reading. If you read the current at a lower-than-normal voltage, your readings will be wrong.

A 1k "shunt" resistor cannot carry the 30mA of normal operation without 30V being
involved - this is a non-starter. bypass the ammeter (or set it to a higher current setting
till ready to make the measurement. Trust me I've done it.