Though I have taught myself how to be better at 'seeing' facial expressions--by studying people in public, and on TV and in films--there are still difficulties. I am quite good now at recognising "happy" facial expressions. But I am still quite bad at noticing neutral or "unhappy" faces. By neutral I guess I mean "subtle", such as the expressions of contentment, relaxation, boredom, or expressions involving slight movements or positions of the facial muscles, smaller nuances. I have to really concentrate if I want to perceive these on a person's face. Most of the time, unless someone is using an explicit and unmistakable facial expression (pure happiness is not hard to mistake, for example) they tend to just look pissed off to me.
I think I could pinpoint almost any facial expression if I had enough time to look at it, but in most situations that's no possible. I think that's where most of the difficulty comes from, my "delayed" recognition of what's happening on somebody's face.
It's like when you're standing at the side of the road and a car zooms past, but you don't recognise it. You just catch a glimpse of it: its colour, the shape of its bumper, the tint of its windows, the sound of the exhaust. Then you construct and image in your head, running through possible car models, narrowing the list down. You finally figure it out just as the car passes over the horizon.