Tuberculosis (TB) is typically attributed to a single infecting strain of the Mycobacterium tuberculosis Complex (MTBC), however mixed-strain infections are more common than reported due to the limitations of conventional diagnostics. While whole genome sequencing (WGS) methods have improved the detection of mixed-strain infections, existing tools struggle to reliably identify mixed-strain infections with frequencies below 10%. TBtypeR, a new tool, addresses this challenge by comparing WGS data to a phylogenetic SNP panel of over 10,000 sites and 164 MTBC phylotypes and using a model based on the binomial distribution to classify sublineage mixtures at frequencies as low as 1%. Extensive benchmarking shows TBtypeR outperforms current methods. Application to a published dataset of 5000 WGS samples identified 305 mixed-strain infections, six-fold higher than previously reported. The TBtypeR R package and a Nextflow pipeline are available at github.com/bahlolab/TBtypeR, providing a powerful tool for studying TB epidemiology and mixed-strain infections.