Well, it is better used because its just easier than actually using sound objects and having to update them all. Its mainly used for big projects, I mostly use the sound object if I want to play a few sounds, etc etc etc. Here's an example of how to use it.

//include the script
#include"sound_pool.bgt"
//define the class
sound_pool p;
void main()
{
//make a few variables
int listener_x=5;
int listener_y=5;
int x=2;
int y=6;
//now we play a sound.
p.play_2d("sound.ogg",listener_x, listener_y, x, y, false);
}

The sound_pool has a method for the river example, but it takes several arguments, and it can be hard to remember all of them and what order they're in.
pool.play_extended_2d("river.wav", listener_x, listener_y, sound_x, sound_y, left_range, right_range, backward_range, forward_range, looping, offset, start_pan, start_volume, start_pitch);
In this case, I'd make looping = true, and ofset and start_ values 0.0, except for start_pitch, because a pitch of 0 is almost inaudible. If you don't want any special pitch, just use 100.
I find play_extended methods to be somewhat cumbersome, so I write a function to save time. However, this works because I usually keep track of the camera separately, and use geometric objects to keep all the position and range values manageable.

Basically, the sound_pool is a fire-and-forget method of playing sounds. It allows you to easily update the entire soundscape when the player moves, but optionally keep variables for the sound slots, if you need to move them individually. You can also set all of your step and behind variables once (or just use the defaults), and never type them again. You can easily pause_all, resume_all, and destroy_all.
You might call it a soundscape manager.