Henry Spencer (whose wisdom is finally becoming available on the web, apparently) explains:
An experienced designer with more freedom to act might have realised that there was just too much optimism in the Ares I concept, that a shuttle SRB was simply too small as a first stage for a rocket carrying the relatively heavy Orion spacecraft. There were several ways to handle the situation, but in my opinion the best was to just forget about Ares I entirely: build Ares V, or something like it, right away and use it for all the launches.
With a big launcher, there would be plenty of margin for weight overruns in development. Using the big launcher for Earth-orbit missions would obviously permit much heavier payloads there. Moreover, the lunar missions would get greater margins too, because they’d be done with two big launches rather than a big one and a little one, so they could weigh almost twice as much.
There is also an important pragmatic issue: the biggest threat to NASA’s return to the Moon is the possibility that Congress will delay or cancel development funding for Ares V. Doing Ares V right away, and using it for the Earth-orbit missions as well as the ones to the Moon, would have ensured that this crucial element of NASA’s plans actually gets built.
Of course, better yet would have been a focus on in-space infrastructure, drawing on ISS assembly experience, to allow us to use existing launchers. That would have also freed up money for earlier development of injection stages and landers, and made lunar missions much more of a fait accompli by now.