"something annoyed him and I'm not sure what" 

Hmm, maybe the "now go and do some math ;)" bit struck him a bit tart but hey, I'm just guessin'. ;) At any rate,

The modular approach usually does end up costing more if you buy all the modules but the rationale is that you're not going to buy all of them. Most people, if apparent use is any indication would be well served with a core that included a basic raytrace renderer. I'd bundle dynamic hair and cloth in one module, the face room in another and an advanced render module as another. The basic render engine would include a stripped down material room and would be able to utilize all shaders but nodes 'n such would be in the advanced module. That's roughly what Cinema 4D does.

From an economic standpoint, if you can get people to pay for features they'll never use, it helps to subsidize those features for the people who do. A lower end version might be better for many users but it would risk cannibalizing revenue needed to develop the more advanced features that are key to where they seem to want to position the application. At some point though, unless your new features are compelling to the majority of users, they may stop upgrading - ask Microsoft.

"Democracy is a pathetic belief in the collective wisdom of individual ignorance." - H. L. Mencken