How are you getting a reading through the OBDII? The stock O2 sensors are narrowbands, not widebands. The only way I know of to get a AFR is with an external wideband, which can be wired up via a firewire cable to feed into the SCT X3 and therefore be monitored by Livewire. That's the way I do it for datalogging.
Unless I'm mistaken, it's a hardware thing. Up until recently, the O2 sensors were narrowbands, not widebands. So even the PCM does not know what the AFR is outside of a very narrow range (+/- a couple tenths from stoich), and even then, it really only knows if it is lean or rich, not an exact value.
OBD2 Makes an estimation. I should maybe say 'most'. I'm not sure. My suburban calculates one and when I visually compare the graphs, there were few anomalies. I use the TORQUE app. No idea how much is calculated through the app and how much though the EEC. The comparison was a tail sniffer set to that specific mode.
I'm sure that digital readouts from each would show more variance. This was only a visual compare of graphs.
I still don't understand how it could "estimate" the actual AFR. There are so many things that can affect that (tip in/tip out, coasting fuel shut off, O2 sensor bias, transient fuel, etc.). The only thing I can figure is that it might be reading the commanded AFR, which is what it's supposed to be, and not what it actually is. But I don't know how the computer could read or estimate the AFR with only a narrowband sensor.
One of the options is to read the commanded AFR. It does take readings from the 02 sensors. It combines them with whatever other inputs it uses to calculate AFR. Keep in min that this is on a stock vehicle. Results could be different if the AFR were to dip into territory outside of the normal range of a narrow band. If the EEC it's doing it's job, then that doesn't happen very often.
Happens every time you go WOT. And WOT is really the only time you should be concerned about what the AFR is, considering the O2's keep it around stoich at part throttle (barring severe mechanical issues). And still, there's no way the computer can accurately estimate the AFR outside of the narrowband range.
It does a better job than you might think. It's not very expensive to experiment with either.
Purchase the app and a bluetooth interface. I think is app is like $5 and the interface $15.