FireLance
Legend
Well, to state my assumptions and arguments more clearly:First, we're talking about 5e (well, really 5.5e) not other editions.
Second, wtf does "exceptional but otherwise typical" even mean?
Third, assuming there is an answer to #2, you cannot logically just jump to a specific statistical interpretation. Even if the language just said "PCs are above average" it could mean they further out on the same curve but it could also mean that they are on their own curve.
1. The range of ability scores for non-PC members of their races follows 3d6 plus racial ASIs. This approximates a normal distribution, and is one way to represent statements such as, "the average gnome is smarter than the average halfling".
2. Assuming a 4d6-1 plus racial ASI generation method for PCs, the PCs' ability scores still fall within the standard range for other members of their race, but are biased towards higher numbers. They are exceptional compared to the norm/mean, but they don't exceed the typical range of ability scores for their race and hence are not "atypical".
3. Using the standard array or point buy without racial ASI PC generation methods still creates PCs with ability scores that are higher than the norm/mean and fall within the standard range for their race (so still exceptional but not atypical), but a 17 Int halfling will be further along the bell curve for his race than a 17 Int gnome. This is not my preferred approach, but I can understand why WotC wants to adopt it.
Your assumptions about the distribution of non-PC ability scores seem to be different from mine, and that is fine. I am just stating my preferences and my conclusions if my assumptions are accepted.