I'd be having a look at the reluctor, to me some of the timing references you mention are actually the sensor not pulling down the 8v supplied by the ecu.
Possibly due to gap problems, can you scope it over a wider timebase and see if its occurring at the same reluctor points each revolution.