Have you chosen the "Sequences from type material" option?

total score usually is better to use, as blast does local alignments - so total score is kind of a sum of over all global sequence. It's clear you have Brevibacterium there - thus you have 1 very good hit to all these bacteria (probably some very conserved gene/protein) - thus max score is very similar .

You can check if in both cases it's +- the same range of you sequence genome.

And then you can check the other 'ranges' for the two matches. Since one has >10 000 total score and another one just around 2000 (with both having the best 1 match >1000) - this means in the first case (with total 10000) - you have many hits there, while in second case (total 2000) - you have the same 1 best hit and probably just very few others.

e.g. your genome is from 1-100, the conserved gene is in the first 1-20 positions. In both cases it matched lets say 2-17 and other case 1-20 positions. Then the 10 000 one also matched 40-60 and 65-100, while the 2000 one could have matched only 55-60, 90-100, positions. Check number of hits and each individual hit to understand this better.

I'd use query coverage and identity rather than either score (or, for that matter, e-value). Exclude every hit below, say, 95% cover, then sort by identity. A cutoff of 97% or 99% is often used as a threshold of intra- vs interspecies variation for 16S, so see what people studying your particular bacterium usually use.

The problem with scores is that they are dependent on alignment length. EG if B linens has an entire genome on NCBI, while only a fragment of 16S rRNA has been sequenced for B siliguriense, the former would a longer match with your query. Not because it is really more similar, but just because a corresponding part of B siliguriense haven't been sequenced. Thus, Best score shows how similar your query is to a certain most-similar site within the reference, while total score multiplies that by a number of homologous sites, which in your case is the number of rRNA operons in B linens (this is a huge oversimplification, but it's basically true).

As for why it isn't sorted by total score when you sort by max score, this is some implementation detail on NCBI's part. Maybe they sort by whatever order they are in memory, or by e-value, or whatever. If you need to sort by multiple columns in a particular order, do so explicitly.

No I didn't used that option Dr. Jakub Grzesiak

Is that option will help to find more accurate results?

Thanks a lot Dr. Ingrida Olendraite , and Dr. Alexei Anatolievich Morozov for your valuable answers.

I still need more information about this point; how to choose the right identification.

If your entire sequence is 100% identical to one and only one seq/species on NCBI, this is most likely your species. If it is 97-99% identical to one sequence and much less identical to others, there is a good chance that this is your species. If it is 90+% identical to a bunch of species without a clear leader, there is no way to be certain, but you may try to build a phylogeny.