Why are total read counts of KrakenUniq different from Kraken2?

Question

[Sarrah Rose]

Hi, I've been running Kraken2 and KrakenUniq (using the --report-minimizer-data flag) on my metagenomic samples. When I sum up the per level read counts for each sample, I get the following values, where it seems like Kraken2 consistently has x2 total reads than KrakenUniq.

As I understand it, KrakenUniq filters reads more than Kraken2. Is that true? Are unfiltered reads stuff it wasn't able to classify? Also, is Kraken2 doing any filtering?

Thanks!