Hi @JvWyatt,
First, I'd like to say that I'm so sorry for your loss! I certainly know what that's like. 
Secondly, these are very promising results! I would suggest altering the --p-min-fold-parent-over-abundance flag of DADA2, and set to 8 or 16, as outline here. This should help increase the non-chimeric values, as those are still quite low. A typical run often sees 70 - 80% chimeric reads. But they do get to ~ 50-60% on poor runs.
Also, I think you'd be able to increase your merges if you set the following, as I mentioned earlier:
--p-trunc-len-f 149
--p-trunc-len-r 149
That last base for each read is quite high and could affect your merges. I'd also perform a simple sanity-check by running dada2 denoise-single. That is, only use the forward reads. This will give an upper-bound on the amount of data you can keep as you will not be dependent on merging. I've had to concede defeat for a few projects myself, where I only used the forward reads. This is totally fine.
I do not think there is a need to re-orient your sequences, given how many appear to have successfully been classified.