Background. The accuracy of dose calculation is crucial for success of the radiotherapy treatment. One of the methods that represent the current standard for patient-specific dosimetry is the evaluation of dose distributions measured with an ionization chamber array inside a homogeneous phantom using gamma method. Nevertheless, this method does not replicate the realistic conditions present when a patient is undergoing therapy. Therefore, to more accurately evaluate the treatment planning system (TPS) capabilities, gamma passing rates were examined for beams of different complexity passing through inhomogeneous phantoms. Materials and methods. The research was performed using Siemens Oncor Expression linear accelerator, Siemens Somatom Open CT simulator and Elekta Monaco TPS. A 2D detector array was used to evaluate dose distribution accuracy in homogeneous, semi-anthropomorphic and anthropomorphic phantoms. Validation was based on gamma analysis with 3%/3mm and 2%/2mm criteria, respectively. Results. Passing rates of the complex dose distributions degrade depending on the thickness of non-water equivalent material. They also depend on dose reporting mode used. It is observed that the passing rate decreases with plan complexity. Comparison of the data for all set-ups of semi-anthropomorphic and anthropomorphic phantoms shows that passing rates are higher in the anthropomorphic phantom. Conclusions. Presented results raise a question of possible limits of dose distribution verification in assessment of plan delivery quality. Consequently, good results obtained using standard patient specific dosimetry methodology do not guarantee the accuracy of delivered dose distribution in real clinical cases.