I think a fairly simple guideline to determine which peaks should be counted in lists such as the Colorado 14'ers, Eastern 600's, and similar situations:
For second-highest: the peak should be the highest that is separated from the true highpoint by a saddle 9/10 the height of the highest peak. (see prominence at http://cohp.org/prominence/index.htm ) Therefore, if peak X was 10,000 feet high, the second-highest peak would require a mandatory climb down to 9,000 feet or lower. If peak Y was 9,500 feet tall, and was connected to X by a saddle of which the lowest point was 9100 feet, then this peak would not be considered the second-highest.
For lists: taking, for example, 14000 peaks. For example a peak 14,500 feet tall would be encircled by a contour (however confusing) of 13,050 feet high. Any other peak inside this area would not be considered a fourteener.
I think this is an easy rule to remember, and would reduce a lot of confusion among those who attempt to climb peak groups such as these.
I can see the logic behind using 9/10 ratio, but applying it to only the higher peak means that any little bump on a descending ridge might be considered a legitimate peak if it happens to be separated by a dip that is 9/10 the height of the main peak. I would think such a requirement should be applied to both peaks.
I typically just use the 300 foot rule--if two peaks are separated by a saddle at least 300 feet lower than either peak, the lower one is a separate peak. Maybe it doesn't have to be 300 feet, but a fixed number does make the math a lot easier.
As far as 300 feet being much bigger on a small mountain than on a large one, keep in mind that because of altitude effects, a 300 foot climb on a high mountain is a whole lot harder than a 300 foot climb on a low one!
A further argument might be that the difficulty of the terrain between the peaks should be considered, but that just gets into a whole other can of worms. I definitely agree that the simplest definition is probably the best!