How do I imitate the sound of a very long tube? How can I waveshape the sound of an overdriven speaker? Where can I download a physical model of the sound of water? What kind of algorithms produce the sound of a Slinky going down stairs (alone or in pairs)?
Is it really so difficult to use the real thing itself? (Of the above examples, the Slinky might be the most difficult thing to locate in one's studio...) *Must* everything be emulated in this imaginary, digital world?
The interesting part is that you can put in some additional variables to change the sound in ways that could never be achieved in the real world. That you can do so without setting up mikes and so on but instead alter the sound in realtime. Which makes it even more useful even if you don't want to go live.
But you are right in that way that natural sounds are much more complex than most of what (pleasing to the ear) we can do in the digital domain. Which is why they usually sound more interesting and not so boring to us. Here some samples, there some live input, can change the world.