I agree, my measurements are probably somewhat (if not totally) out of whack with the mic in my hand, the speakers on the desk, and the AC running in the background. I'd like to learn how to take better measurements but I haven't really come across anything that describes how to do it very well.
The crossover schematic I have in the first post contains the 3rd order butterworth filter in addition
-the DC resistances of the inductor windings (the resistor in series with the inductor coils on the schematic),
-the L-pad for a 3.5 dB tweeter attenuation in the high pass network,
-impedance compensation (the capacitor and resistor in parallel) for the rising impedance of the woofer's voice coil at higher frequencies,
-and the modeled DC resistance and inductance of each drivers voice coil along with its mechanical response (the right most portion of the high pass and low pass networks)
So you're right, it's not a pure 3rd order butterworth filter, but it's in there.
The parameters used to model the impedance presented by the drivers at different frequencies were derived using this website: http://www.shine7.com/audio/orcad.htm
The site describes how the drivers can have their impedances modeled using the DC resistance of the driver's voice coil along with its inductance in addition to the back EMF calculated by the driver's moving mass. The program he uses is Orcad/Pspice, which is a circuit simulation program that electrical engineers (like myself) are very familiar with. So instead of modeling the crossover with a constant 8 or 4 ohm load, you can substitute that model in and get the exact response of your filter to the changing impedance of the driver.
How the values for the model came to be I'm not sure about. But what I do know is that once calculated, the results are strikingly similar to what's measured from the driver in real life.
Using the tweeter I used, you can model it like below:
You can then perform an AC sweep and then dividing the voltage by the current running through the circuit you can derive the impedance of the driver:
You'll find that it looks strikingly similar to the published specs from the manufacturer:
All that's left to do is to plug that impedance response into filter/crossover you'd like to model and you'll get a close to real life transfer function of the filter with the particular driver across the frequencies of interest.
It makes sense to me as an engineer, I hope that clears things up for you.