I decided I needed to double check the O2 sensor voltage calculations that use the Nernst equation instead of trusting some internet graphic, because I remember thinking this wouldn't work in the past...with complete combustion, the sensor would only be at 0.1v if you're just 1% lean.
So the issue is incomplete combustion and how much O2 is left in the exhaust stream. In this paper, it cites this graph from a 1987 paper that I can't find on Google, which seems to suggest 0.5% at stoich!:
https://www.researchgate.net/figure/..._fig3_34286046
Of course, exhaust is hot so some of that is probably consumed by the time it hits the O2 sensor.
So narrowband closed loop fueling doesn't target stoich, it targets the point where the engine is running rich enough to consume that excess oxygen by the time the exhaust reaches the sensor. So all narrowband controlled engines run just slightly rich, but changing the switching voltage doesn't represent a meaningful change in oxygen concentration until you drop it past the point where the sensor is useful.