Output watts per tube

I originally posted this question in the amp/pre-amp section with limited response. Thought I'd try it here, with some corrections. I'll further preface this by saying that I am not a technical guy, so everything I'm telling you here I learned last Thursday - fair warning if there are any gross inconsistencies with my explanation.

I recently visited a guy locally who was able to test the dozen EL34 Ruby output tubes from my Cary V12, with interesting results. Despite having 6-800 hours on them, ten of the twelve tested between 100 and 110 on a tube tester, and the other two came out between 85 and 100. This was using the sensitivity set at 53, per that tester's standard for the tube type.

Then, we used a sine wave scope with three tubes to get an average control measure, then checked all of the tubes against the control for distortion (none) by running two tubes in a test amp running push-pull in class A. This was also hooked up to a meter to measure output watts. (someone asked, and I don't know what the plate voltage was used in the amp) The result was that those two tubes - any combination of two of the twelve - had a sustained maximum output wattage of 21 watts, even the two that were a little low on the tester. The results from tube to tube were, unexpectedly, almost perfectly linear. Given the Rubies less than stellar reputation, I was expecting some measurable degradation or at least inconsistency.

So here's my questions: Is the 21 watts for two tubes the equivalent of 10.5 output watts per tube; and if my amp takes twelve tubes and makes 50 watts per side in triode, am I in fact only using a little over 8 watts per tube at any given time? The amp can also run at 100 watts per side in ultra-linear, which would require closer to 17 watts per tube (six per channel). Does the 50 Watts equate to relatively low stress on the tubes and longer tube life? Does employing 12 tubes somehow reduce the per-tube output further? (Cary didn't have an answer for any of this)

Any thoughts?
There are a number of good tube data manuals on the internet. Here are the numbers from Duncan:

Vh - 6.3
Ih - 1.5
VaMax- 800
Vg2Max - 500
VhkMax - 100
PaMax - 25 (watts)
Pg2Max - 8
IkMax - 150

In short, running full-tilt, you can get 25 watts per tube. Many amp makers run tubes under their maximum ratings as that helps tube life and makes it easier for the amp to meet specs over the long haul. For example, the classic Dynaco Stereo 70 got 35 watts out of pair of EL34s.
I am by no means an expert, but I'll give you my line of thought for an answer.

The maximum value for the EL34 of ~25 watts is for pentode operation. You can take a bit more than half of that number for the maximum output in triode operation (~13 watts). The Stereo 70's I've seen that run in triode have pulled between 8.5-10 watts per tube. As I understand it, triode mode isn't actually easier on your tubes in general, it just utilizes fewer of the internal elements. I think your amp is probably biased to a point that is about normal for current tube amp designs.
So apparently the amp that we were using as a test bed was also not running the tubes anywhere close to their maximum output capability either. Based on Jazzerdave's estimation, the 8 watts per tube that the V12 is using now is about 2/3s the power the tubes are actually capable of producing. OK, that helps clarify it. It's more of an academic point, but it was still interesting to see.
Despite having 6-800 hours on them, ten of the twelve tested between 100 and 110 on a tube tester, and the other two came out between 85 and 100. This was using the sensitivity set at 53, per that tester's standard for the tube type.
this number refers to the "mutual conductance" (a.k.a. transconductance) that the tube delivers in the tester's circuit, which is a representation of the change of plate current divided by the change in grid voltage. With output tubes, this translates to "power gain" (power output vs. AC grid voltage), not maximum power output before clipping.

Maximum power output before clipping in a tube amp is a function of the plate voltage, the tubes' plate resistance (Rp), the load impedance (the speaker load impedance multiplied by the square of the output transformer's turns ratio), and the output transformer's insertion loss. All of these factors except for Rp are part of the amp, not the tubes . . . and this tube parameter tends to be less variable than transconductance.

Also note that "Pa max 25 watts" does NOT refer to the maximum power of an amplifier using this as an output tube, rather it refers to the maximum amount of power the tube's plate (a.k.a. anode) is able to continuously dissipate. (Pa = "anode power"). Plate dissipation is a function the quiescent (idle) conditions, the loading, and the duty cycle. I've personally measured almost 220 watts output from a quad of "25 watt" EL34s, running push-pull pentode with about 730 volts on the plates.