Without proof or independent validation, I don't accept the lifespan data provided by any manufacturer as real-world.
Fair enough. I've done a fair amount of searching myself, and there doesn't seem to be any independent validation that is available to the public. I've often wondered about this myself. Over the years, I've considered wiring one in parallel with an hour meter just to check, but never gotten around to doing it. Even if I did, it would be anecdotal rather than statistically significant.
As I understand it, the claimed "lifespan" (e.g. 8,000 or 10,000 hours) is MTBF.
Roughly, this means that at the predicted hour marker, 50% of the units will still be in operation. In the case of linear fluorescents, this applies to the lamp only, not the ballast. IEC 60969
is the operative standard for integrated-ballast CFLs in the EU, and covers the entire unit. Certainly, misapplication will negatively affect life expectancy. I know that nearly all CFL failures I have seen have been ballast failures.
As for the bad from stock scenario, I don't know if any data is collected on the number of BFS CFLs. I know it is far less than 25%, however!
Let's look at a parallel from the world of toilets. Many of us will recall that the Energy Policy Act of 1992 mandated that toilets sold in the US after 1996 could use no more than 1.6 gallons (6.0 liters) of water per flush. Most manufacturers hastily modified existing designs to use less water, with unsatisfactory results. Although this problem was largely fixed by 1999, it took much longer for the reputation of low-flush toilets to recover, and some people still will not modify their opinions of low-flush toilets no matter what new evidence comes to light.
Until recently, there was no way to evaluate the performance of a particular model of toilet other than hearsay, advertising claims, and the meager lists published by a few water utilities of models they had had good luck with.
Now we have a credible, unbiased, objective guide to which toilets really work. It is called the MaP study,
and it's worth a read, if only for the hilarious detailed description (with photos!) of the "test media". If we had that for CFLs, it would settle a lot of debates.
As for the embodied energy
of CFLs, that info is not publically available, either. The best you can find is a few educated guesses.
They may be higher (and possibly better educated) than my guess, but they don't constitute proof.
Let's assume Rod Elliott's worst case estimate of 20kWh to manufacture a CFL, vs. 1kWh to manufacture an incandescent. Let's also assume the manufacturer's claimed MTBF, 1,000 h for incandescent and 8,000 h for CFL:
0.015kW* 8,000= 120kWh
0.06kW* 8,000= 480kWh
20kWh* unit CFL= 20kWh
1kWh/unit incand.* 8 units= 8kWh
Clearly, if the MTBF numbers are reliable, CFLs are the winner. The question is whether (and in what situations) they are reliable numbers. That is why independent scientific studies are needed.