Don't quote me on this, but when you're working with mixed mode (part analog, part digital) circuits it's generally best to keep digital parts separate from the analog parts. What I mean is having separate grounds, supplies and such is a good idea. This is why the PT2399 has an analog ground pin and a digital ground pin. If you have common grounds etc. you'll find that noise from the digital parts of the circuit will creep into the analog parts. So yes, there is a reason why you shouldn't use the 5V as the bias. Also, what's the point? How much will 2 resistors and a capacitor cost?
In general I agree about the cost and space concern for components, but I try to keep things as small as possible to fit the circuit in a small box. In a small build, 2 resistors and a capacitor cost a lot of valuable real estate. But more importantly is that two voltage traces becoming one greatly simplifies the layout. When you have traces for ground, signal, 4.5V, 5V, and 9V things can get hairy, especially in a single sided tight layout.
I see what you are saying about keeping analog and digital references separate. However it is reported that the analog and digital pins for the PT2399 should both be connected to analog ground. You can read this on Merlin's site (Small Time description and the PT2399 notes). I don't see how using a 5V regulated reference as a bias is any different than taking a 9V regulated reference, dividing the potential and using it as a bias.
