here are two examples.
This first exposure is utter garbage. The second photo is after I tweaked it in post. The skin in the second image is right where I like it. Now I just want to take the second image in camera!
Clearly the E-3 was underexposing the man's face here and I had to do some considerable lifting to restore it.
In this image I'm guessing you were using ESP metering - the E-3 averages points from all around the frame and meters to an average of those averages. You can see that the upper right 1/4 of the image is extremely bright - the E-3 takes this into account, and will likely underexpose this image. I would expect the E-3 to underexpose in this situation, assuming you metered either on the face or metered exactly as composed.
How does one "nail" the exposure like the second image when you don't have a chance for two shots. In this case are we talking about AE bracketing?
I can think of a few different ways to achieve this, however I'd simply use ESP (since I'm so familiar with it) and apply positive exposure compensation. This is really a matter of preference and experience, though I have to admit I probably wouldn't "nail" the exposure the first time around on this image using this method, despite how confident I am with ESP metering. This is partly because I rarely take these types of images, and partly because I try and refrain from it. Wedding photography is different, and requires you to do this, so perhaps one of these other methods will work.
One method is to use ESP+AF metering mode. This is the same as ESP, but puts emphasis (averaging) on the focus point used, rather than giving all focus areas equal priority. I have little experience using this so I can't confidently comment on it, however if you're very familiar with ESP, this may be the first thing to check out.
Secondly, and perhaps the most straightforward answer, is to use spot metering. As you can imagine, spot metering averages a very small spot and sets exposure based on that. In theory this works well for uniformly lit subjects, however it would not work well in the scenario you linked to, unless you understand how spot metering works. For example, if (in a hurry), you spot metered on the man's right side of his face (left side of the picture), you would get a very different exposure than if you spot metered on his left eye (right side of the picture). You'd probably get exposure differences of at least 1.5EV, which is huge. If you happen to accidentally spot meter on a hot spot on his/her face, it gets even worse.
The "fix" for spot metering is center-weighted metering, which puts priority on the focus point, but also takes into consideration the immediate area around the focus point, as well as (to a much lesser extent) the remaining portion of the frame. However, I'm not sure if center-weighted metering works for any focus point other than the center one, which of course would make it almost useless. You'd have to check the other focus points.
Every situation is so dynamic and constantly changing. I can't set it and forget it.
This is why I'd suggest choosing a metering method that works for you 75% of the time, and then figure out how to work around with "fixes" for the remaining 25% of the time. In my case it's easy, since I'm confident with ESP metering in about 90% of the scenarios, and have the option to adjust for the remaining 10%.
Also something interesting to point out, the camera is trying to make fair skin 18% grey which I don't ever really want. And dark skinned people need a bit of negative EC applied in camera. I see images of people over at fredmiranda and also here at DPreview where the skin tones are spot on and exposure is amazing in each photo. Are they just pp'ing them to correct initial issues from the RAW file or are they that good OOC?
Note that the camera isn't trying to make the skin 18% grey, but it's trying to make the portion of the image that is metered to be 18% grey. If the face isn't metered, it may not be 18% grey. If you meter a solid black box (with a person standing on it), the camera will attempt to make the solid black box 18% grey, and in the mean time the person will be greatly overexposed.
Remember, no one ever posts images that are sub-par. Why would you? We post images that are near-perfect, or appear to be. So third parties look at those images and often mistakenly believe "all" their images would be that way. In fact, I'm guessing a lot of people rattle off hundreds of images before they post one onto DPR or FM.
--
Tim
'I haven't been everywhere, but it's on my list.'
E3/7-14/12-60/150/50-200/25/25/EC-14
http://www.flickr.com/photos/timskis6/ 
