How to Improve Functional Gene Varaint Calls?

Hi all,
I intend to demystify two cases in my analyzing.
1-unlike the 16S rRNA gene analysis, if I increase the identity percentage with both Blast+ and Vesearch classifiers, the variant calling is slightly going better! Why? And my bar plot indicates many unaligned taxa. I checked out taxonomy reference extracted from NCBI by Rescript plugin in Qiime2. Many lineages are uncultured while the project I am doing expected more known taxa based on the published papers. I double-checked steps from importing to DADA2 as well as merging and grouping steps! No any problems detected! I am suspicious to the database (sequence and taxonomy references) which obtained from NCBI. Now, 2-I need to improve the result so your guidance is required.

Please share your idea @SoilRotifer

Supplementary info is below:
Qiime2 version: 2020.08.
The functional gene: mcrA or methyl coenzyme-M reductase A or methyl coenzyme-M reductase Subunit A

86% identity.

90% identity

Taxonomy reference:

Many thanks