Sure, maximum power/loudspeaker goes down as you connect more speakers in parallel due to power dissipation in the amplifier.
However... I would have expected the voltage gain of an audio amplifier to be independent of load impedance. Therefore, I would expect the output voltage of the amplifier (and the output per speaker) to remain constant as speakers are added in parallel... at least up until the point where we hit the power dissipation limits mentioned above (or the amplifier loses stability when the load impedance gets too small).
Where am I wrong?
Assuming an infinite power source from the wall, amps are limited in their output in two ways - voltage and current.
The power supply rails limit the maximum voltage the amp can produce. This limitation comes into effect when the load impedance is high - usually more than 8 ohms - the maximum voltage the amp can produce within the limits of the supply rails delivered to a high impedance is less than the rated output of the amp. Because of this, increasing the load impedance - putting drivers in series, super high impedance drivers, etc., will reduce the power the amp puts out compared to a more normal 8 ohm load.
The more common limitation in real world use is current limiting. Most amps run into current limiting with 4 ohm loads, and virtually all are current limited driving loads with less than a 4 ohm load. In this situation, the low impedance means that the current limit is reached with an output voltage less than the maximum rail voltage. How this manifests itself is that the output voltage of the amp stops rising at whatever value it hits when the amp's max current output is reached.
Different amps have different compromises. The ITechs are fairly high voltage amps, which means they can make a lot of power into 8 ohms for their nameplate capacity. The trade off is that their 2 ohm performance falls off significantly compared to some other designs with more current capacity - they have almost the same power at 2 ohms as 4 ohms, where a more conventional design would have significantly more power available at 2 ohms than at 4 ohms.
Power dissipation inside the amp is potentially a limiting factor (and is certainly related to the current supply limitation of the amp), but the amp's internal impedance is much lower than the series impedance of speaker cabling plus drivers. For example, an amp with an internal impedance of 0.2 ohm driving a 2 ohm load with 1/2 ohm of speaker cables is a total impedance of 2.7 ohms. A "much better" amplifier with an internal impedance of 0.1 ohms into the same load is a total impedance of 2.6 ohms. Even though the amp is dissipating half the heat as before, the total power into the circuit isn't much different. See discussions of the meaninglessness of damping factor as an amp specification for more info.
Also, if you have high impedance loads, you're going to be limited by the voltage rails, not the heat lost in the amplifier.
There are further nuances depending on how the amplifier is set up - an AB amp actually dissipates the most power at zero output, and is fairly efficient when run wide open. The various multi-rail designs like class H and other variants means that there are fractional power levels where the amp is quite efficient, and other fractions where the amp is significantly less efficient, depending on where your signal level falls relative to the rail voltages used in that particular amp.
The other significant real-world limitation is the stiffness of the wall power to the amp. Even modest power amps can draw 50 or 60 amps for brief periods when they are charging the supply rails. Though this isn't enough sustained demand to trip a 20A breaker, Ohm's law still applies, and the input voltage can sag significantly, reducing both the voltage and current available to the amp.
Apologies if my logic and/or math is slightly off - it's past my bedtime.
