Why? I honestly don't see how that's self-evident for single-player games. Hell, even at the highest level of multiplayer competition, Starcraft 1 (designed for interesting gameplay first) ended up a lot better than Starcraft 2 (designed for balance first).
Because design of complex systems is a complicated. You end up with a lot of subsystems that interact with each other and changes of one subsystem can (and often will) have unforseen consequences throughout the whole system. Introducing even simple changes to a system in the middle of its implementation can force you to redesign (and thus reimplement) a big parts of the system, often it can even make it more cost effective to trow away your whole work and start from scratch.
Because of that before you even start designing a system you have to clearly define problems and goals and (ideally) make a process where design and implementation phase are completely separated. Analogy that is often used is building a house, it is just more wise and easier to first make the blueprints and then build than to simply start building and then demolishing every time you want to change room layout.
No matter what, you are going to end up tweaking some numbers, but if you try to solve problems you didn't think about during design more often then not you are going to end up with a disaster.
That's all nice and true. However, designing interesting and complex systems is difficult. My point is, designing them while also focusing on balance is extremely difficult, because in many ways, "interesting" and "balanced" are at odds when it comes to gameplay mechanics. The more complex and fun something is, the more balance problems it results in. So, the focus on balance, in my experience, leads to mechanical homogenization and dumbing down, and the removal of options which are deemed too troublesome from a balance perspective. I know many, many examples of this, although they're mostly from multiplayer games, where this kind of thing is a lot more prominent. SC2 vs SC1 is probably the most high-profile, but there's a lot of others.
And you know, multiplayer games are between a rock and a hard place here, because they need to be as balanced, and as fun as possible, at the same time, and very, very few get this completely right. What I don't get is why anyone would willingly tie this boulder around his neck when designing a successor to IE games, which you yourself admitted were imbalanced as fuck, and everybody loved them anyway. This is true for a lot of the Codex' most loved games, actually. Fallouts, PST, Arcanum, System Shock 2, to name a few, are heavily imbalanced as well. And it honestly doesn't matter that much (well, it does in Arcanum's case to an extent, because you have to larp harm not existing).
I joked a bit earlier in this thread that Sawyer is designing single player, party-based WoW, but you know what's scary? WoW also went through several phases of streamlining and homogenization, where they attempted to balance PvE and PvP at the same time. And you know what? Some of their ideas at the time were eerily similar to Sawyer's. Funnily enough, D&D 4E was also compared to WoW for similar reasons.
Honestly, I expect PE to be a fun game that I will enjoy a lot. But I'm not holding my breath for Sawyer to prove himself a systems design savant and getting it right immediately. It would be nice if he did, but not expecting it.