Mapping assembled genome back to reads: why SNPs?

[bio_informatics]

Hi members,

I've paired end, Illumina, Klebsiella Pneumonia.
Read length 251, MiSeq run
Quality remains good until the read end.

Assembled genome's quality is reasonable checked with Quast.

When I map my reads back to assembled genome, and carry out SNP analysis, I find SNP within it.
I set coverage as 10 to call a base SNP.
I fail to understand why a reference has SNP when mapped against itself.

Is it something to do with SPAdes' correction?

Tool used for mapping, and VCF - bowtie, and GATK.
spades version: 3.5.0

Spades run on auto coverage settings, everything else default. Bowtie, and GATK default.

I found similar post at:

http://seqanswers.com/forums/showthread.php?t=41739

Any help shall be appreciated.