In college I took a class on CFD. It was enough to make me really interested in CFD, but not enough to become an expert.

Anyways my prof said that you should basically always use a pressure boundary instead of an outlet. Her reason was that with a pressure wall the flow can recirculate back into the fluid domain where as with an outlet it can't. If the fluid really would recirculate in this area then you should not constrain it not to (in fact I hear it can make your solution diverge).

I was wondering if anybody had another opinion on this. I designed an air intake duct for an engine and using the pressure boundary the flow did indeed show some recirc. I think in reality this would not happen because the intake would probally pull in the air before it can recirculate.

Maybe I need to use a pressure boundary with vacuum pressure instead of atmospheric pressure?

Your professor is right, when you are now boundary condition. But sometimes you can't make experiments. For example third step in 8 steps turbine. In such model you are don't know the outlet from second step. So you must to use outlet, because you didn't know what number you should give for pressure parameter.

I would have to disagree. While a pressure boundary lets flow back in, that flow will almost certainly have the wrong velocity, turbulence and temperature values. The code cannot model flow outside the domain.

In general backflow at a boundary is bad. If it is a long way downstream of the area of interest it may be acceptable on a 'good enough' basis. It implies that the domain is wrong, you have cut the model in an area where the flow is not 'simple' and in an area where you cannot know the boundary condition accurately. You will need to extend the model, in your case probably to the the gap where the flow enters the cylinder past the valve. Here you have a high probability that the flow is not recirculating.

Outlet conditions are useful as the flow propagates to the exit boundary and the BC allows a spatial variation in pressure and velocity while maintaining a mass flow (which you may well know to a reasonable accuracy). This property makes them widely applicable and robust.

There are however times when, for example, the pressure really is constant and then they are obviously not correct.