That would be the best way to measure the total energy stored in the battery, but as the article points out, it's necessary to take samples at regular intervals to get an accurate measure of this.
A simple runtime test with a fixed value load doesn't fully measure this, since the voltage (and hence current) both decrease as the batteries are being discharged. In other words, the number of watts is decreasing with time, although with NiMH probably not by too much. He says he takes data at regular intervals, then integrates (sums) his data over the runtime to get the total watt hours. Actually, he doesn't explicitly say this, but I know what he means.
It's also necessary to consider that batteries deliver power with different efficiencies under different loads. A larger load produces less efficient power delivery (I assume because of excess heat). So his fixed value resistor set up may not tell the whole story, depending on how "draining" your application is.
Then there's the factor of draining the batteries in spurts instead of continuously. It's a real consideration for digital cameras, but for high power flashlights too, since most people switch them on and off.
I can't tell you anything about these other factors.
