How so?

I assume that some smart algorithm could make this a whole lot less painfull (by eg trying to analyse sentence/paragraph/word structure of the selected area) or, simply, show a popup with a zoomed-in version of the selected text, allowing for simple & quick modifications. Or something else aiding selection, I don't know. The point is - I'm positive that some creative thinking "outside the box" (yeah, I hate that overused phrase just as much as you do) will lead to better usability than we have right now (as you admit yourself it's less than stellar).

"Not-so-perfect" hardware is very subjective, if you ask me.
Is the current NIT touchscreen hardware anywhere near perfect? Not to me, to be honest. I'd much prefer having a screen that reacts reliably to the slightest fingerbrushes and traces my fingers accurately while moving them across the screen (and that is impossible with a resistive TS), even with the disadvantages that brings. Because most of those disadvantages can be worked around with smarter software, while you can't work around something the system doesn't even register like unnoticed TS contacts which are so common with resistive screens.
And for the "magically" better apps: the vast majority of Maemo software is created by enthusiasts who begin programming for the NITs because they want to have that certain app for themselves, first and foremost. They're constantly eating their own dogfood (to abuse yet another overused phrase), so to speak, so they have every motivation to create a usable UI. And that will happen despite limitations like finger-only TS usage and ultimately lead to a much more consistent UI/UX, which is a good thing in my opinion.