How could it be that there is a recirculation zone near the region defined as outlet (i.e. no flow through the boundary)?

Outlet boundary is a numerical boundary which assumes zero gradient (except pressure) in the flow direction. Thus it has to be chosen in a place where velocities and scalars have little change. If the position is not chosen properly, then you will have recirculation around the boundary and a converged solution would be very difficult to obtain and the results can not be trusted. Try extend your solution domain or use pressure boundary.

John, thank you for your advice. Can you tell me what options should I apply to pressure boundary in order to model the normal atmosphere conditions? I've tried static pressure = 1e+05 Pa, and Environment and Mean options On. Is that right?

For Incompressble flow, the pressure is only relative since the total mass flow is fixed by inlet boundary condition and pressure at inlet will be adjusted to satisfy the overall mass balance. For compressble flow, the value of proper pressure is important. For your case, you should be OK with your specification of pressure, usually environment pressure option is advised if you have inflow /outflow in a large opening.

It is not advised to have both Environment and Mean options On. And don't forget that the value of pressure is always treated as relative to reference, so if you have reference pressure of 1e+05Pa you should specify 0 at the boundary.