Thanks for the additional detail cjones - I'm learning something here too. So, now you've made me start thinking more about it.
What's the need for the upstream injection when the engine is cold? Perhaps when cold without the O2 signals, it's running a bit richer, and in that situation, all three cat chambers need additional O2, hence the injection further upstream? I know that the toughest part of the passing the EPA tests was getting through the 'cold' start up portion. If messing with the O2 sensor readings is truly at issue, I wonder why they didn't simply didn't arrange the plumbing to inject downstream of the O2, but upstream of the cat; seems that plumbing would be a lot easier/less expensive than the crossover tube, casting the heads with the air injection passages, machining them, etc.