All you need to do is come up with a V/I curve of the LED's at the currents you measured and use that diode drop calculated against your equivalent MA comparison and you'll get a true efficiency comparison, all it would take is to plot the diode drop with the initial calibration currents you did.
Some of the Joule Thieves I made ran as high as 600kHz. You will need special gear to measure the current drawn at this frequency, as the system behaves like a low power factor load.
I don't think power factor is a problem because it's DC voltage but it is true that you cant use a DC DVM to measure the current because the current is AC. You could use an oscilloscope and calculate the RMS current from the waveform.
The filter circuit of page 4 filters out completely the AC ripple component. I verified the voltage drop across the 0.47 ohm resistor with an oscilloscope and it is essentially DC.