In any point buy system of any sort (whether for attributes or whatever) there needs to be a recognition that the incremental value of an additional bonus in a particular thing increases at rate that is faster than linear.
This is because it is much better to be overwhelmingly good at something than it is to be mediocre at alot of things. In all RPGs and RPG inspired games, breaking the game involves finding a way to synergistically due one thing extremely well - often to the exclusion of ever doing anything else. This is itself a synergy. If the one thing that you do well, you do so overwhelmingly well that you never have to do anything else, then you don't have to 'waste' any resources getting good at anything else.
So pretty much any RPG worth talking about does something to penalize overly focusing on one thing. In D&D's attribute point buy system it is primarily by weighting extraordinary increments as being more valueable than less extraordinary ones. This is correct design.
Similarly, we know that a +5 sword isn't worth merely 5 times a +1 sword. A 20th level character isn't merely 20 times as good as a 1st level character. A +6 bonus to your spell DC's isn't merely 6 times better than a +1 bonus to your spell DC's. And so forth.
Frankly, I find the plea to stop weighting extraordinary scores more valueable to be just a bit disengenious. I don't think anyone making that suggestion really expects to use the new system to create a wider variaty of ability score arrays. Quite the contrary, I think that they - and most players - would be strongly discouraged from producing wider variaties of arrays.
If 4e is valuing an 18 even more highly than 3e, then that is something 4e is getting right. Compared to 3e, you are much less dependent on being good in a wide variaty of abilities. Compared to 3e, in 4e you have much less penalty for having low scores in anything. It's already looking like an optimal array for 4e would be along the lines of 18, 18, 12, 8, 8, 8 - and that you'd happily drop those 8's lower if you could. If the 4e designers recognized that they reduced some of the penalty for min/maxing and compensated, then kudo's to them for that at least.