I played with this recently and not sure if this answers your question…
Channel mapping presented a 7.1 or 5.1 DTS bitstream to my HT processor which locked it in DTS with sound coming from 2 speakers. Downmix properly sent a 44.1khz (or whatever) 2-channel LPCM stream to the processor which then let me decide how to process that bitstream in the processor. I’d say, if you want 2-channel stereo use Downmix.
For my set-up it was better to downmix as I didn’t like sending the 2-ch audio as DTS encoded to my processor. I preferred sending via LPCM. YMMV and may be dependent on your processor. If you’re using a 2-channel DAC I’d downmix. But, nothing stops you from trying both ways and listening for yourself. It may very well be environment dependent on what sounds better.
I assume you have a two channel setup and a multi-channel source file.
Most likely downmix is is going to have ‘all the music’.
With channel mapping you may miss some of the mix, some mixes put vocals on the center, which would get dropped. Roon explanation of multichannel is here.
Another option is to do your own mix, making a combination of the above. For example map the left & right channels to their normal speaker, center & LFE to both left & right and drop the surrounds. Don’t forget to add some headroom (or adjust each speaker level accordingly). Here is a screenshot of that using DSP procedural EQ and Mix: