For example, if you have only potential values of 100 to 900 with an expected average of 500, you get a bell curve in a Benford analysis. But, if you expand that range to include double digit numbers, and single digit numbers and maybe four digit numbers...then the Benford curve starts taking affect.
No it doesn't. That's the point. If a random variable has a Gaussian distribution (bell curve) then the measured values of that random variable will not follow Benford's Law, no matter how many orders of magnitude are spanned by those values.
The issue is that you have to determine the of the influence of Benford. If you have a system that has a bell curve within a magnitude, then Benford has no effect.
"Influence of Bedford"? That's a meaningless term. If the data follows a logarithmic scale, Benford's law works, and a lot of data does indeed follow a logarithmic scale. If I take five dice, roll them, and multiply the values together, the result will follow Benford's Law, with or without any "influence". If I take five dice, or five thousand dice, and add them, I will get bell curves, and the data won't follow Benford's Law, because bell curve values don't follow Benford's Law. Logarithmic values follow Benford's Law.
If you have that same bell curve and expand possibilities beyond that magnitude, then Benford starts having an effect. The degree of that effect is determined by the possibilities of expansion beyond that magnitude.
No it isn't. Beford's Law will be apparent only in data that follows multiple orders of magnitude, but that is a necessary, not a sufficient, condition.
(Actually, it isn't even necessary. If I generate uniform random numbers between 1 and 10, and take their logarithms, the result will follow Benford's Law, but in naturally occurring data sets, it's unlikely such a thing would ever occur that it is logarithmically distributed without varying over several orders of magnitude. If you have such a data set, you can change the units so that the data does vary among several orders of magnitude, and we're back to Benford.)
And the dominance can change for each number. That is why in my calculation, Trump mostly followed a Benford curve, except bent higher on lower numbers and had a weird bump at three. And Biden had mostly a bell curve but was bent higher at lower numbers and was way up on one.
I would love to see that analysis, because I'll bet it was a lot like the analyses that were published in the blogs shortly after the election. Every data set with lots of 1s, and fewer 2s, and not many of anything else was declared to be Benford's Law, but those analyses were simply wrong. If you have more 9s than 8s, that isn't Benford's Law, even if you have lots more 1s than either, and it isn't anything overlaid with Benford Data.