Student's t-distribution is used for the direct estimation of the modeling and geometrical perturbations in the Monte Carlo simulation of fusion blankets. A test of hypothesis is carried out for the equivalence of the means for the reference and perturbed systems at different confidence levels. If the test is failed, intervals for the difference of means or perturbation can be directly deduced. No variance reduction is attempted in the application of this methodology. Application of the methodology to the neutronic and photonic analysis of the conceptual HYFIRE high-temperature process heat fusion reactor blanket is carried out. The use of a two-dimensional model for the analysis versus one-dimensional models leads to differences in the estimated system parameters (e.g., breeding ratio) ranging from 1.5 to 7% at the 70% confidence level. Accounting for the penetrations, using three- versus two-dimensional models, affects those system parameters in the range of 12.8 to 20.9% at the same confidence level. These uncertainties are judged significantly large and need to be accounted for in future reactor designs.