You are misinterpreting this. The E-M5/1 does not care for the focus area at all when it detects a face. It focuses on faces anywhere inside the frame, including outside the focus area. The focus area is only used when it does not detect a face (in time).
Now that "(in time)" is the important part. In order for the camera to focus on a face in AF-S it needs to detect the face *before* you half-press the shutter button. A face being detected is shown in a *white* frame!
Unfortunately the camera *does* keep detecting after focusing begins and may thus detect a face when it comes more into focus during AF motor movement. Even when it does not focus on that face it will *expose* for it in ESP/matrix metering mode! A face being exposed for is shown in a *green* frame!
When white and green frame are on top of each other is the only time you get both, focusing and exposure on a supposedly detected face. This is why falsely detected faces can be a problem and why FD is better turned off when not needed. This may also improve focusing times in general as the processor has less work to do.
Using AF-C can help, because then the camera then corrects its focus if a face is detected after first half-pressing the shutter button. This may take longer, though, because when a face is not detected before the camera will again first focus on the focus area and only then correct itself on the newly detected face.
If you have the time then it's best to pre-focus close to the face to help the camera detect it "in time". This is mostly a problem of fast aperture lenses and close distances where depths of field is small.
--
Red flash eyes save lives and eye-sight!
http://en.wikipedia.org/wiki/Retinoblastoma