Heat transfer simulation result is much larger than experimental result

I modelled the heat transfer through a hollow brick to calculate the equivalent thermal conductivity. So general heat transfer and weakly compressible Navier-Stokes modules were taken to build the model. But the simulation result is a little larger than the value obtained from experiments. The result of simulation is 0.91 w/(km), while experimental result is 0.68 w/(km).

I consider there should be some problems with my model (probably convection), but I can't find it. The simulation is successful, no error messages. So what could be the problem causing the big difference between computational result and experimental result?

Can anyone please help me?

Thanks in advance
Xiqi