See, I used to think this until I heard a talk by Kendall Miller in which he described his pricing experience.
He had an enterprise tier that was orders of magnitude more costly than the other tiers. Executives would skip straight over the other tiers and buy the enterprise level happily. Customers would come to him and say "I love you guys! You really understand the enterprise."
That tier did have extra features, just like you're encouraging. But the punchline is Miller logged the use of every feature and discovered not a single one of his enterprise customers was using the extra ones. They simply wanted to pay more.
His response was something similar to what you said about not feeling the need to make every bit of money possible. The Basecamp prices felt right to him and he was comfortable leaving money on the table to stick with pricing that matched his gut feeling.
That's not exactly what you were talking about, but I liked his way of thinking.
Also, do you have any data for a lower price = happier customer = more referrals who pay?
That's a lot of assumptions built into that premise, each one of which could be tested.
If we're going to pretend this is all a priori knowledge and not something that can & should be tested, then let me try!
Here's mine: "People hate to feel like they're having every penny rung out of them… so they probably love buying the middle tier and enjoy laughing at the silly people who would pay $100 extra for something they don't need." Works just as well as your theory.
If you're going to have different pricing tiers, the higher tiers need to ostensibly have more work put into them. You can't just say that if you fit into one bucket of people we know you have more money, so we're going to charge you more (e.g. DVD regions). You need to give them something more for it, so that your customers don't hate you (e.g. HD movie downloads).
Nobody likes to feel like every possible penny is being wrung out of them, or that they're being unfairly targeted to pay more. Measuring happiness of customers or how likely they will refer others is more difficult than measuring which product each customer bought, but just because it's harder to measure doesn't mean it's less important.