Sorry Bill but all of the real world evidence says otherwise. I'm well aware of Keele's work on that subject but there is no violation of Keele's mathematics at work in car audio SPL You are just assuming there is. The efficiency noted in the article is 25% for a direct radiator but that is only based on nominal power input not true efficiency or input.
Quoted example below.
"This relatively low value of absolute maximum efficiency (rather than 50% which would be
expected for a load connected to a generator with the load resistance equal to the generator's source
resistance) is a direct result of the definition of nominal electrical input power used in the definition
of efficiency. If the input resistance of the driver were defined as being ?.RErather than RE, as it is
in the definition of efficiency for horn loaded drivers , the maximum efficiency would be 50%
rather than 25%, as it is here. Alternately, efficiency could be defined as relating true output power
versus true input power (rather than nominal input power), which would remove this limitation and
would allow efficiency to rise to 100%. Unfortunately, this would mean that the efficiency
frequency response function would differ from the response function measured with constant input
voltage, and thus the efficiency frequency response would differ greatly from the response under
normal operating conditions. As an example, the true efficiency of a closed box system at box
resonance can approach 100%, because the input impedance could be very high at resonance, thus
making the input power quite low. This would reflect as a peak in the efficiency frequency
response that would not be representative of the frequency response as the system is normally
used, with a constant voltage drive source.
This power transfer condition is exactly the same as a generator having a source resistance of
R1 driving a load of resistance R2. Maximum power is transferred when the load resistance is
equal to the source resistance. In this case, however, the efficiency is reduced by a factor of 2,
because the input power is twice as high due to the definition of nominal electrical input power."
Additionally not only is this a nominal approximation of input power but he is talking about efficiency as a % where you are making a jump to maximum SPL levels possible but have not defined what environment this data point occurs in , which directly impacts it, the distance from the radiator,etc. This has a direct impact on the SPL that will be measured and is why there is such an increase inside of a vehicle. The speaker is enclosed inside of a vessel that is very small compared to the wavelengths of interest. The environment not only modifies the sensitivity of the device but also the theoretical nominal SPL/ efficiency ceiling that you speak of. It is not the same SPL at 10m outdoors as it is at 4m or 1m is it? Why would it be the same inside of a small car? It is most definitely not. Looks like the inside of my jeep allows something near 128-130dB theoretical 1w sensitivity. No laws are being broken.