Why I got more reads after the quality control of KneadData?

I use KneadData to handle a pair end metagenomics data, but it turns out I got more reads after using bowtie2 to remove the contamination.
I download the database by using: $ kneaddata_database --download human_genome bowtie2 $DIR

The package I use is installed form conda:
kneaddata 0.7.4 bioconda
bowtie2 2.3.5.1 bioconda

It’s wired I got more reads after quality control, I would like to know how to fix this.

Here is the command and output:
~$ kneaddata -i /home/~/Lab/MetaBGC/TestData/20samples/ERR2608606/ERR2608606_1.fastq.gz -i /home/~/Lab/MetaBGC/TestData/20samples/ERR2608606/ERR2608606_2.fastq.gz -o /home/~/Lab/KneadData/ERR2608606-2 -db /home/~/Databases/KneadData --trimmomatic /home/~/Biosoft/anaconda3/share/trimmomatic-0.39-1 -t 26 --trimmomatic-options “SLIDINGWINDOW:4:20 MINLEN:50” --bowtie2-options “–very-sensitive --dovetail” --remove-intermediate-output
Decompressing gzipped file …

Decompressing gzipped file …

Initial number of reads ( /home/~/Lab/KneadData/ERR2608606-2/decompressed_crkn7up2_ERR2608606_2.fastq ): 19868184.0
Initial number of reads ( /home/~/Lab/KneadData/ERR2608606-2/decompressed_vi3caoel_ERR2608606_1.fastq ): 19868184.0
Running Trimmomatic …
Total reads after trimming ( /home/~/Lab/KneadData/ERR2608606-2/ERR2608606_1_kneaddata.trimmed.1.fastq ): 18157607.0
Total reads after trimming ( /home/~/Lab/KneadData/ERR2608606-2/ERR2608606_1_kneaddata.trimmed.2.fastq ): 18157607.0
Total reads after trimming ( /home/~/Lab/KneadData/ERR2608606-2/ERR2608606_1_kneaddata.trimmed.single.1.fastq ): 1444354.0
Total reads after trimming ( /home/~/Lab/KneadData/ERR2608606-2/ERR2608606_1_kneaddata.trimmed.single.2.fastq ): 132765.0
Decontaminating …
Running bowtie2 …
Total reads after removing those found in reference database ( /home/~/Lab/KneadData/ERR2608606-2/ERR2608606_1_kneaddata_hg37dec_v0.1_bowtie2_paired_clean_1.fastq ): 33873371.0
Total reads after removing those found in reference database ( /home/~/Lab/KneadData/ERR2608606-2/ERR2608606_1_kneaddata_hg37dec_v0.1_bowtie2_paired_clean_2.fastq ): 3788405.0
Total reads after merging results from multiple databases ( /home/~/Lab/KneadData/ERR2608606-2/ERR2608606_1_kneaddata_paired_1.fastq ): 33873371.0
Total reads after merging results from multiple databases ( /home/~/Lab/KneadData/ERR2608606-2/ERR2608606_1_kneaddata_paired_2.fastq ): 3788405.0
Total reads after removing those found in reference database ( /home/~/Lab/KneadData/ERR2608606-2/ERR2608606_1_kneaddata_hg37dec_v0.1_bowtie2_unmatched_1_clean.fastq ): 229142.0
Total reads after merging results from multiple databases ( /home/~/Lab/KneadData/ERR2608606-2/ERR2608606_1_kneaddata_unmatched_1.fastq ): 229142.0
Total reads after removing those found in reference database ( /home/~/Lab/KneadData/ERR2608606-2/ERR2608606_1_kneaddata_hg37dec_v0.1_bowtie2_unmatched_2_clean.fastq ): 0.0
Total reads after merging results from multiple databases ( /home/~/Lab/KneadData/ERR2608606-2/ERR2608606_1_kneaddata_unmatched_2.fastq ): 0.0

Final output files created:
/home/~/Lab/KneadData/ERR2608606-2/ERR2608606_1_kneaddata_paired_1.fastq
/home/~/Lab/KneadData/ERR2608606-2/ERR2608606_1_kneaddata_paired_2.fastq
/home/~/Lab/KneadData/ERR2608606-2/ERR2608606_1_kneaddata_unmatched_1.fastq
/home/~/Lab/KneadData/ERR2608606-2/ERR2608606_1_kneaddata_unmatched_2.fastq

I think my recent post may be relevant. [Kneaddata FASTQ header problem]