JR and JBM, I'll try a little more sassifictition. The frequency spec question was discussed by Ian and others in
this thread a couple of years ago and may be of some interest. In relation to the -9dB bass falloff point, it's also interesting to note that in a paper by Dr. Toole, who researched these matters at the NRC with his colleagues(including Ian and Alan), he uses -10dB below the 300-3000Hz average as being the best figure for usable bass extension. The =/- 3dB spec is simply what is commonly adopted for describing speakers.
As to the sensitivity specs, since the walls of the anechoic chamber at the NRC reflect essentially no sound to be added to the direct sound entering the microphone, in a typical listening room the reflections add to the sound level which was measured anechoically. This varies with the liveness of each room, of course, but a 3dB increase is often suggested for a typical room. A related aspect is that although loudness drops 6dB for each doubling of listener distance under anechoic conditions(e.g. at 4 meters it would be 12dB lower than at the 1 meter distance at which the sensitivity was specified)in a live listening room the decrease would be more on the order of 3-4dB per doubling of distance and the power required for a comfortably loud listening level wouldn't be increased as much(probably in the area of 2 watts would do).