Gold standard for Mitochondrial DNA variant calling

Question

[Anh]

I'm developing a variant calling pipeline for mtDNA data generated from long-range PCR following by NGS. I want to compare my pipeline with different pipelines for mtDNA variant calling by designing performance metrics (precision-recall curve, F1-score, ROC, and AUC) but I need a gold standard variant truth set (reference data with label variant or not for 16569 base pairs of mitochondria) to evaluate performances among pipelines. Can you suggest comparing pipelines should I follow that approach and whether my approach to comparing pipelines should follow which gold standard?