1. What res are you playing in and outputting to?
2. Have you tried using DxTory's "Distribution Writing" feature to lower lag?
3. Why do you need to split audio streams?
The res you record games at, esp power hungry games, can make a big difference in performance, as well as the res you output to.
Distribution Writing involves designating 2 or more drives to simultaneously write the capture file, but you have to use the DxTory codec for this feature. It's not a bad codec, but writes a VERY large file. When I use it I only use it in the Low or Med mode, and only record moderate length sessions, then set it to auto delete so I don't overfill my drive. The DxTory codec also requires you to use RawCap output mode vs AVI. The RawCap files then need to be joined by the RawCapConvert tool built into DxTory, which is where the auto delete option comes in., You can set it to auto delete the RawCap file after the AVI is constructed. I also delete after each clip is recorded. The Low setting should be near equal to Shadowplay on max, but still probably larger file size.
If you only want to add background music or a commentary audio track, you can demux the game audio with something like VirtualDub, then put the game audio track in a tool like Audacity, and add a music or commentary track to it, then put the joined audio file back in. It's imperative the audio track you add is equal or shorter length though.
I would use Shadowplay in a heartbeat if I had a GPU that supports it. Main reason being the sheer convenience of being able to play and replay segments until you get footage you know you want to keep, and just click it on afterward. It avoids all the hassle of drive clutter and trial and error recording.
So, in summary, consider the above options carefully and do some tests to see what is best time and difficulty wise for you. It may at first seem like Distribution Writing is better, but when you consider the file size and time it takes to reconstruct files, demuxing and mixing audio tracks starts looking just as good if not better, esp considering you have the convenience of ShadowPlay.
What I really liked about Audacity is I could custom fine tune the tracks, such as making the background music lower volume during key dialog scenes. You could do the same with commentary, but maybe in reverse, lower game sound a bit to emphasize commentary at certain points. There's a lot you can do with audacity, even add volume fade, change pitch, etc.
https://www.youtube.com/watch?v=W0KqsTa_HpM
https://www.youtube.com/watch?v=3uqCNjbQn54