After reading Tvad's question in the post just above this one, I ran my own estimates of the power required at an 8 foot listening distance. Assumptions are a semi-reverberant room, and we're listening to stereo (both channels driven).
Anechoically, sound pressure level falls off by 6 dB for each doubling of distance from a point source. Eight feet is about 2 1/2 meters, so we could expect a falloff of about 8 dB relative to the 1 meter SPL. But we have two speakers, each driven by its own separate amplifier channel. Double the total cone area equals +3 dB, and double the total amplifier power equals another +3 dB, so we gain back 6 dB right there. And this is a semi-reverberant room, so we can expect a dB or two of room contribution. Let's be conservative and estimate 1 dB additional from room contribution at 8 feet (at greater distances, the relative room contribution would be more). Plugging in the numbers, we have -8 dB distance falloff, +6 dB for double the cone area and double the amp power, and +1 dB for reverberant field contribution, for a net change of -1 dB compared to the single-speaker 1 meter SPL. So back at 8 feet, from a stereo pair of 89 dB/watt efficient speakers we could expect 88 dB SPL at the listening position for a 1 watt input. Similarly, with 94 dB efficient speakers, we could expect 93 dB at the listening position.
Elevick's chart is for a single speaker at one meter, but as you can see it's certainly in the ballpark for what you can expect from a stereo pair at your approximate listening distance. One thing the chart necessarily leaves out is the effects of power compression, but that's a whole nother can of worms and can't be generalized in a such a chart as it varies significantly from one speaker to another.
Now Gs5556 brings in the issue of peak-to-average SPL, and he's right in estimating 15 dB of headroom being required to avoid clipping. On some recordings the peaks can be as much as 25 dB above the average level! Fortunately, the ear is relatively forgiving of modest amounts of clipping. In fact, many popular recordings have numerous "flat tops" (clipped waveforms) already. The extra power requirments figured into his numbers are to avoid clipping on instantaneous peaks, but if you don't have that much power available doesn't mean you can't still enjoy the music. You're just missing a little bit of the liveliness and sense of effortlessness that having no amplifier-induced clipping would give you.
At least, that's my take.
Tvad, the fact that you're running a tube amp helps a great deal if you're pushing the amp into clipping sometimes. Tube amps "soft clip" (relatively little high order harmonic distortion produced), while solid state amps "hard clip" (lots of high order harmonic distortion produced). The ear is vastly more sensitive to high order harmonic distortion than to low order harmonic distortion. You've perhaps heard something like "one tube watt equals two solid state watts"? Well, the reason has to do with the audibility of the type of distortion they produce when driven into clipping.
Duke