Ugh, just lost half an hour of typing. I'll try again.

I've never delved into a one power discussion before, but as a stats teacher I thought I might be of some use.

To the above, and especially the bolded - yes the first statement is correct, assuming that's what RJ said, but it has nothing to do with the second statement. The second one is fallacious.

An example given above had a mean of 1 and a standard deviation of 4.1, but that was a little abstract, so why don't we use a real example? The average male height in the US is 5 foot 10 or so. That doesn't mean that the maximum possible male height is 11 foot 8, no one has ever even come close to that. All that it means is that if you took all the males in the US, added up their heights, and divided by that number of males, you'd get a mean of 5 foot 10. Shortest or tallest has no relation to that mean, in and of itself.

The second component to a bell curve is standard deviation, sort of like an 'average spread' of the curve. The standard deviation is around 3 inches (according to a quick internet search) for male height. According to the way that normally distributed data are distributed (the bell curve follows certain rules, and most things in nature are normally distributed, which makes it a heck of a lot easier to analyze stats), that just means that 68% of all males will be between 5 foot 7 and 6 foot 1 (+ or - 1 standard deviation from the mean). But even if you're 3 SD's above the mean, or even if you're the tallest man in the world, it doesn't relate to the mean at all, any more than any number or place on the normal curve does.

(note: of course, with dwarfism there would actually be a little bump in the lower part of the real distribution of height, but let's say we're talking about 'height of people not affected by a physical condition' )

In this power example, then, Lanfear can be ten times stronger than the average Aes Sedai, a hundred, whatever - ratio isn't relevant to a normal curve. Normal curve only talks in terms of probability, so if Lanfear was 3 standard deviations above the mean, it would mean that there's only a .15% chance of someone being stronger than her in the population (as 99.7% of all scores fall within +/- 3 SD of the mean, and the rest has to be split on both ends, meaning .15% below 3 SD below the mean and .15% above 3 SD above the mean). My guess is that Lanfear is even rarer, maybe 4 or 5 SDs above the mean. Regardless, that is meaningless (no pun intended) with regards to her strength relative to other channellers. She could be 1000 times as strong as the channeller that sits perfectly at the middle of the distribution, or 1.5 times as strong. She can still fit perfectly fine into a bell curve of channellers.

With regards to the problem of zero - practically, that's not so much a problem. As I've already mentioned, only .15% of the population falls below 3 SDs below the mean, and that number gets exponentially smaller with each SD you get away from the mean. Taking height as an example, then, as I said you're at 5 foot 1 when you're 3 SDs below the mean. You still have 20 more standard deviations to go before you get to zero, so by the time you get to zero, although there's practically a CHANCE that you could get someone who has zero height, the chances of that happening even assuming it was physiologically possible would be essentially one in infinity, close enough. So the fact that there are physiological limitations doesn't mean that a significant part of the population has power (or height, or whatever) slightly above zero and then hits a wall. For the most part, the zero cutoff is a non-issue in measurement, which is why it's perfectly acceptable to describe things as conforming to a bell curve, even when some of the extreme aspects of a bell curve aren't perfectly replicable in real life. I'd say RJ would be totally justified in describing power as fitting a bell curve, without having to qualify 'except for specific theoretic mathematical points that make very little difference in reality'.