The standard-used Taq DNA polymerase is not a proof-reading polymerase and prone to replication errors (long known among those who cloned a lot, hence, we changed to proofreading polymerases quite some time ago) It even says so on Wikipedia: "One of Taq's drawbacks is its lack of 3' to 5' exonuclease proofreading activity[4] resulting in relatively low replication fidelity" with reference to the original paper: Lawyer FC, Stoffel S, Saiki RK, Chang SY, Landre PA, Abramson RD, Gelfand DH (May 1993). "High-level expression, purification, and enzymatic characterization of full-length Thermus aquaticus DNA polymerase and a truncated form deficient in 5' to 3' exonuclease activity". PCR Methods and Applications. 2 (4): 275–87. doi:10.1101/gr.2.4.275. PMID 8324500.

To confirm that the heteroplasmy is genuine (which it may well be), one should repeat the experiment with a proofreading polymerase to exclude PCR artefacts.