The FR curve you're showing is not because the tested load is 13 ohms but because it's 9 ohms at some frequencies, 70 at others, 90 at others, 20 at others, and so on: so it's variable.
You won't get frequency response variation from a source with high output impedance into a load with constant impedance over frequency, just the worse damping ratio. That's why I showed the graph of the AH-D2000 impedance over frequency.
It's pretty clear that you get a very simple voltage divider circuit between the load impedance and the source output impedance. V_L = V_s * Z_L / (Z_L + Z_s). Voltage across the load, at any frequency, will be the source voltage multiplied by the ratio of the impedance of the load at that frequency to the sum of the impedances of the load and source. If Z_s is constant over frequency (generally it is), then Z_L varying over frequency means you get varying V_L over frequency. Hence wonky graphs like you linked. If Z_L is constant over frequency, that means that you lose voltage to the source output impedance, but it's an equal amount at different frequencies, so there's no relative shift in treble or other frequencies.
Edited by mikeaj - 12/27/11 at 1:40pm