ABSTRACT Low-level rifampin resistance associated with specific rpoB mutations (referred as “disputed”) in Mycobacterium tuberculosis is easily missed by some phenotypic methods. To understand the mechanism by which some mutations are systematically missed by MGIT phenotypic testing, we performed an in silico analysis of their effect on the structural interaction between the RpoB protein and rifampin. We also characterized 24 representative clinical isolates by determining MICs on 7H10 agar and testing them by an extended MGIT protocol. We analyzed 2,097 line probe assays, and 156 (7.4%) cases showed a hybridization pattern referred to here as “no wild type + no mutation.” Isolates harboring “disputed” mutations (L430P, D435Y, H445C/L/N/S, and L452P) tested susceptible in MGIT, with prevalence ranging from 15 to 57% (overall, 16 out of 55 isolates [29%]). Our in silico analysis did not highlight any difference between “disputed” and “undisputed” substitutions, indicating that all rpoB missense mutations affect the rifampin binding site. MIC testing showed that “undisputed” mutations are associated with higher MIC values (≥20 mg/liter) compared to “disputed” mutations (4 to >20 mg/liter). Whereas “undisputed” mutations didn't show any delay (Δ) in time to positivity of the test tube compared to the control tube on extended MGIT protocol, “disputed” mutations showed a mean Δ of 7.2 days (95% confidence interval [CI], 4.2 to 10.2 days; P < 0.05), providing evidence that mutations conferring low-level resistance are associated with a delay in growth on MGIT. Considering the proved relevance of L430P, D435Y, H445C/L/N, and L452P mutations in determining clinical resistance, genotypic drug susceptibility testing (DST) should be used to replace phenotypic results (MGIT) when such mutations are found.