benoit - Hmm, sometimes this site blows with the posting business! I wrote a nice paragraph or two about all this but it got 'lost'. Frankly, I cannot do it again, no enthusiasm now..... 
 
 
Any, remember that the AB design confirms that the 25 ohm resistor is always in place in the circuit when any attenuation is engaged. This means that, looking at the numbers, the first 'click' of the AB is speaker only, the next click - and all others - have the 25 ohm resistor in parallel with the speaker load. Using the math to determine the actual speaker  impedance then:
25 X speaker impedance (plus the added resistances at each click, .i.e. 6 - 11 - 16 - 25) ), divided by the value of the parallel resistor plus the speaker impedances (X, X+6, X+11, X+16, X+25) = load in theory.....
We get, with the 8 ohm cab:
8 (speaker only) - 8.97 - 10.79 - 12.24 - 14.22
In practice (from original posting):
7 (speaker only) - 8.5 - 10.2 - 11.7 - 13.7
Here's the 16 ohm cab in theory:
16 (speaker only) - 11.7 - 12.98 - 14.03 - 15.53
In practice (from original posting):
13.7 (speaker only) - 10.9 - 12.2 - 13.3 - 14.8
Well, OK, so benoit, I get that you question the actual readings vs. the theoritical impedances but you can see that they are pretty close. And I still would suggest that the 16 cab is a better match for the box but eh, you sure won't hurt anything with the higher load.  

Most people stall out when fixing a mistake that they've made. Why?