My recollection when considering 5nm vs 3nm for Ha filters was that the NII was more significant in planetary nebulae and Wolf-Rayet bubbles. So for emission nebulae like M42, you're not going to see much benefit or difference. You need to make the comparison on planetaries like M27.
Sounds like you've done some research, so am sure you've read the section on pages 4 and 5 here:
http://www.astrodon.com/uploads/3/4/...rowbandfaq.pdf which is what I based the above comments on.
For me it was a no-brainer, 5nm for Ha and 3nm for the OIII and SII and be done with it. Could not see any benefit of the 3nm Ha for practically all cases.