To not derail the GL2000 thread, I brought you here 
Looking for a simple answer to a simple question, that may sound right.
Problem is two fold:
-
Depending on the inner workings on the amp (I wrote an overview of all but the Op-Amp based ones here), you may have a linear power to load impedance, except for two power spikes into 50 and 200 Ohm.
-
Voice coils, Ribbons and Planar drivers do not present a linear impedance. In fact, the impedance may be a mountain range similar to the frequency response.
Example:
Source
Don’t get me wrong, guestimating the SPL an amp can get out of a speaker/headphone using a spreadsheet is perfectly valid.
As you said yourself, loud ≠ quality
The Correct-ish Method
I have tried to measure power into a headphone.
Setup was less than ideal as I had a lot of probe leads dangling around acting as antennas, I have no LCR-meter on hand to verify AC-characteristics of the precision resistor I had in series with the drivers and the Oscilloscope I have on loan is a bit slow on the FFT.
The results were as expected: Shite
Learned a lot in trying though.