Hi David,
Well, it depends what you have in mind as a plan. There are options galore. I would recommend some perusing of the documentation (even the 1.7 user manual pdf will cover compose module well enough) and user notes for background. The user manual is also full of handy tips and special techniques.
To your specifics though - the separate stacks would need to be registered, even if you plate solved to the same spot for acquisition. I would also compose your duo band into R, G, and B. No reason to leave out the B, there should be data in there!
I am uncertain if LP filtered data would add anything useful as L, even if you also used one of the options to include a synth L from the duoband. My guess is that the LP data will be noisier, require stronger Wipe settings, and really by and large will only be "seen" in those areas where your duoband channels are adding color.
Though perhaps Ivo or others can confirm if my thought processes here are on the right track or off the rails.
If it were me I would instead probably just rack up more integration time through the duoband and compose as bicolor from OSC/DSLR.
Or, if your LP filtered datseta actually seems reasonable, despite some missing bandpasses, you might consider composing it as R, G, and B, whilst then loading the duoband set into NBAccent.
But again it probably all depends on what the datasets are like and whether the target is properly amenable to composition in this way.