Different flagstat number after using MergeBamAlignment

Question

[ThomasLam]

Hi everyone, I'm utilizing the GATK Best Practices for analyzing mitochondrial NGS data.

I utilized FastqToSam to generate a ualign BAM file. Following that, I aligned the fastq with BWA MEM. Per the Best Practices, I executed MergeBamAlignment with the ualign BAM and aligned BAM. I observed that the original BWA MEM BAM's read1 and read2 counts differ from those after MergeBamAlignment.

BWA MEM BAM flagstat

984290 + 0 in total (QC-passed reads + QC-failed reads)
14478 + 0 secondary
0 + 0 supplementary
0 + 0 duplicates
938859 + 0 mapped (95.38% : N/A)
969812 + 0 paired in sequencing
484906 + 0 read1
484906 + 0 read2
905466 + 0 properly paired (93.37% : N/A)
921184 + 0 with itself and mate mapped
3197 + 0 singletons (0.33% : N/A)
0 + 0 with mate mapped to a different chr
0 + 0 with mate mapped to a different chr (mapQ>=5)

MergeBamAlignment flagstat

1004409 + 0 in total (QC-passed reads + QC-failed reads)
14471 + 0 secondary
0 + 0 supplementary
0 + 0 duplicates
938578 + 0 mapped (93.45% : N/A)
989938 + 0 paired in sequencing
494969 + 0 read1
494969 + 0 read2
892314 + 0 properly paired (90.14% : N/A)
900542 + 0 with itself and mate mapped
23565 + 0 singletons (2.38% : N/A)
0 + 0 with mate mapped to a different chr
0 + 0 with mate mapped to a different chr (mapQ>=5)

Both read 1 and read 2 of the input FASTQ contain 484906 reads each.

Could someone explain the increase number (from 484906 to 494969)?