OK, so the problem here is that you can set up fairly sophisticated routing for headphne mixes and effects while recording so on via Sonar itself, but that means introducing latency. Depending on your computer and the complexity of the project, this may be low enough to work with; it's certainly worth exploring.
Using direct monitoring instead is possible, but with some limitations.
For direct monitring, you only have three output options, "Main", "Sub" and "Digital 1". These are all stereo.
You can't route individual inputs to different outs for monitoring, they're just either on or off.
I've done a fudge for recording singers with reverb in their headphones a few times. I have their mic coming in on one channel, and the direct mix outputs for both Main and Sub on. The singer is listening on headphones to just the Main output.
I then patch in a hardware reverb (I just use a cheap pedal for this myself) to the sub output and bring it back in on another input channel. The reverb is set to fully wet as you would for a bus effect. This then also apears in the singer's headphones.
You've probably already realised the problem with this, which is that the reverb signal coming back in will also be going out of the SUB output and back to the reverb and into an infinite loop. In parctise, I find that with low levels of reverb, I do ok, and it doesn't go into a big feedback howl.
The other obvious limitation here is that you end up with monitoring reverb on everything. I haven't figured out a way of doing it ore selectively.
So there you go; bit of a cludge, but it can be useful.
To be honest, for the scenario you describe, I would try to get your latency low enough and do it through software. That way, you have access to far more outputs, and can set up different headphone mixes. You don't really have that in direct monitoring (you can do two different headphone mixes, but that requires using the Sub output, so you can't do it at the same time as my reverb trick).