The best way to think of it is getting a horizontal slice of the world.   To be quite honest, the camera is a highly visual concept, and I have a feeling that it can't be explained with text.  If I had a demo, I have a feeling it would be more obvious.  Think of it a bit like this: you, the viewer, are watching your character, and you can change where you're watching the character from.  In the design in my head, the camera's position is fixed, unless the player moves it.
That said, expanding the scope of the project and just making it first-person might actually be doable, though that requires a ton more math and wouldn't allow me to really do what I really want to do: a hack and slash rpg.  When you're in first-person and get swamped with enemies, it becomes impossible to really do anything, which is why I want to avoid that.
To fix the HRTF problem, it is only necessary to make it obvious that you are on the same horizontal plane, above, or below the sound.  The distance calculations will take care of the rest.  Actually making the HRTF work for distances closer than one meter is difficult and not appealing enough to get someone to do the research.  If someone has an Anechoic chamber, high-end recording equipment, and someone sighted to read some papers, it would be possible to do it.  You basically need a second HRTF dataset for distances closer than one meter, and at that point the fact that we all have slightly different head and ear shapes shows up more (there's huge datasets of HRTF data for different ear shapes, and anything I write would allow the advanced user to swap out the data for that reason).  You take a dummy head and actually play sounds, collecting something called the impulse response.  This is the point at which the fact that I've not taken DSP classes means I don't understand it.  One day I will, when we have good OCR for math, heh.