From what I've gathered the general consensus seems to be that long tube headers move the narrow band O2 sensors from their original locations in the stock manifolds. This has a couple of 'side effects'.
First, when you move the NBO2 away from the cylinder, it doesn't get as hot as the stock location which skews the NBO2 voltage slightly. Here's a graph of the difference, I can't speak to the accuracy of the test data but it gives you the idea.
3_narrowo2_afr.gif
Secondly, the ECU uses the Closed Loop Airflow and Closed loop integrator delay to adjust the amount of fuel injected to induce NBO2 switching. When you change the location of the NBO2 the time it takes from injection to response at the NBO2 is different and thus confuses the ECU into thinking the fuel isn't responding and it adds or subtracts more fuel.
You'll likely notice that the LTFT and STFT are all way off the VE you would expect if tuned with your wideband.
This is as much as I've gathered in my reading. Take it for what it's worth.