An extensive set of measured data was developed for the purpose of verifying the accuracy of a photon dose‐calculation algorithm. Dose distributions from a linear accelerator were measured using an ion chamber in a water phantom and thermoluminescent dosimeters in a heterogeneous anthropomorphic phantom. Test cases included square fields, rectangular fields, fields having different source‐to‐surface distances, wedged fields, irregular fields, obliquely incident fields, asymmetrically collimated fields with wedges, multileaf collimator‐shaped fields, and two heterogeneous density cases. The data set was used to validate the photon dose‐calculation algorithm in a commercial radiation treatment planning system. The treatment planning system calculated photon doses to within the American College of Medical Physics (AAPM) Task Group 53 (TG‐53) criteria for 99% of points in the buildup region, 90% of points in the inner region, 88% of points in the outer region, and 93% of points in the penumbra. For the heterogeneous phantoms, calculations agreed with actual measurements to within ±3%. The monitor unit tests revealed that the 18‐MV open square fields, oblique incidence, oblique incidence with wedge, and mantle field test cases did not meet the TG‐53 criteria but were within ±2.5% of measurements. It was concluded that (i) the photon dose calculation algorithm used by the treatment planning system did not meet the TG‐53 criteria 100% of the time; (ii) some of the TG‐53 criteria may need to be modified, and (iii) the generally stated goal of accuracy in dose delivery of within 5% cannot be met in all situations using this beam model in the treatment planning system.PACS number(s): 87.53.–j, 87.66.–a