This paper deals with the relevance of Computational Fluids Dynamics (CFD) results confronted tomeasurements carried out under uncontrolled thermal conditions. Experimental tests have been undertaken in a room and in a kitchen of an experimental house. Although the wall surface temperatures and the air intake temperature have not been imposed, the air change rates have been controlled during the measurements. Moreover, since measurements have been carried out in a real environment, air leakage has occurred at the walls. Thereafter, measurements have been used to define boundary conditions of CFD simulations.As a result, it can be stated that as long as air leakage is slight during the experimental tests, satisfactory agreement is observed between CFD results and measurements.However, in the case where experimental tests are carried out in a room with significant air leakage, the quality of the numerical results has been decreased since this phenomenon is not considered in the CFD model. For this case, a simplified hypothesis of modelling has been proposed and validated to deal with air leakage effects and thus to improve the accuracy of CFD results.