Kraken and Centrifuge have close results, please refer also to this blog https://www.sevenbridges.com/centrifuge/

I like centrifuge better since you get the number of reads that are uniquely classified to a certain genomic sequence. This is likely to be more specific. So if you are doing comparative genomics, this might be what you want. MEGAN is great too, however, it is slower compared to the other tools. But it provides functional analysis of datasets which is beyond simple taxonomy.