Gravar-mail: Overcoming bias and systematic errors in next generation sequencing data