I know what you're saying, but I think the marketed sensitivity has some utility.
A car's "City miles per gallon" rating is considered highly inaccurate and unattainable in any realistic driving, But I consider it at least a benchmark for the best the car could achieve (if they advertise 20 MPG City, I can bet that I will get worse than that, but most likely not 25 MPG).
Likewise, if I think an application needs a minimum of 95dB sensitivity, I feel like I can drop the speakers with < 95dB from the running. Like, I wouldn't worry about investigating the 87dB speakers.
The problem between your comparison of sensitivity and the MPG on a new cars sticker is this:
The cars are measured in the same manner, so while the number may not reflect what you will actually get, the relationship between each car basically remains the same. With speaker sensitivity, it is not measured the same by every manufacturer. lets take for example a speaker that measures in a chamber 87db one watt per meter.
if that speaker is a 4ohm speaker a manufacture could list it as 90 watts per 2.81 volts.
Another manufacturer could use in room measurement and list the speaker as 94db one watt,2.81 volts.
Now we are talking about the same speaker, but showing 7db difference. Some manufacturers measure using two speakers and some corner load a speaker. So without knowing exactly how the measurement was taken, listing sensitivity means very little.