No, when you think about it it's less processing involved than using point camera.
In the end it's a personal preference thing, I like the drift, but if I have the camera positioned higher than the object, I'm effectively shooting over the top of it, though my camera is tracking it. If I rotate the camera say 20 degrees down, I'm once again pointing at the object, and following it. The xrotate command can be before the loop, so it's actually less processing going on.
To demonstrate this, go to the sliding collision example, take the line that reads: 'xrotate camera 15'
and place it before the do loop.
It remains rotated, without having to be processed every time.
I used to do what you suggest, and use point camera, but it really is quite processor hungry compared to the above method.
ZX Spectrum 48k Issue 3, Radio shack Tape drive, Rank arena 12" T.V. set.