The problem is the focal length. When you look at the image using the orthographic views, in this case the front view, the focal length is infinite. This doesn’t exist in reality because it would mean a scene where the perspective lines never converge. Imagine a pair of train tracks. If they are seen in a orthographic view, the tracks would never converge.
The shorter the focal length, the faster the tracks come to a point in the distance.
That is why when using a macro lens, everything looks distorted and objects in the distance seem way smaller and farther away than the foreground objects. If you use a telephoto lens, like the ones photographers use at sporting events, the objects in the distance look a lot larger than in real life and closer as compared to objects in the foreground.
If we apply this knowledge to your image, it’s easy to see part of the problem you are having. (The other part has to do with anatomical issues, but I won’t deal with that here.)
When you look at your work through the perspective view, the ears which are farther back seem to get smaller and the seem to disappear behind the zygomatic arches. The sides of the head converge to rapidly as if looking through a fish-eye lens. On the other hand, the nose feels enlarged because the heightened perspective makes it disproportionately bigger as it comes closer to the lens. The shorter you make the focal length, the larger the nose will be in relation and the smaller the ears will be.
When you look at your image through an orthographic view, There is no perspective (this doesn’t exist in reality), the focal length is infinite. So as a result, the ears aren’t affected by perspective and as a result they look larger than they would appear in real life. The nose is affected by feeling smaller than it would in real life. You can test this by opening the image in a program like maya and sliding the focal length of the camera back and forth while maintaining the image at screen size. This is what Alfred Hitchcock used to do to create a scary uncertainty in some of his shots. They called it a Hitchcock Zoom. He would zoom out while dollying the camera in and closer to the figure. This would create that crazy horror movie effect where the person’s face remains the same size while the room seems to get larger as all the background elements seem to get smaller. Check out the movie “Vertigo” to see this effect in full bloom!
The human eye has a focal length that is somewhere between 50 and 75. A professional portrait photographer, who is worth his salt, will use this knowledge to make the people in his portraits look better. Many will use a telephoto lens in the studio and place the camera farther away in order to do close ups. This makes the people look beautiful because it makes the nose look smaller than in real life. Usually any focal length above the human eye’s focal length will generally cast the close-up of the human face in a favorable light.
Because you are using reference that has perspective, while working in a view that has no perspective, You are not getting accurate results. Since there is no foreshortening on the ears and nose in your mesh, you are fooled into thinking that the ears are smaller than they are. But in actuality you are making the ears too small. The reverse is true of the nose and it results in the nose being too big on your mesh. Thus the distortion that you are getting. The perspective view of you model is a more accurate representation of what your mesh looks like in reality.
In order to work around this problem you need to do one of two things. First of all, you must not work in orthopedic views. You should lock your cameras in perspective views. Do this for all your views, front, back, side, top, and all the 3quarter views. If you are working from image planes, you must do your best to match the focal length used in the reference shot. This provides most accuracy. If you aren’t working from image planes, you should set your camera focal length to somewhere between 50 and 75. Even if you are not using reference, it’s a rare occasion where it is better to work on a portrait while looking through orthographic views. Imo, I suggest you don’t do it at all when working with heads.
The best 3d portraitists, usually only use image planes set up on and looking through perspective views. This is usually for the blocking in stage and to double check proportions. They mainly use the image plane (in locked perspective views) to capture the primary forms of the head. Then they open up as many reference pics of the model that their monitor will allow and eyeball it from there as if they were working from life. This way of working allows for the most accuracy.
I suggest you take your reference shot and attach it to a locked perspective camera in front, side, and 3quarter views. Then get the main shapes that way. And if you are going to resolve the portrait from the image plane reference, then at least stop using the orthographic view to work from, at least when you are doing faces.
(Btw, the lack of a decent perspective camera and the lack of a focal length setting in earlier versions of z-brush is partly why the skymatter guys created mudbox when they were at WETA.
Personally, it used to drive me crazy when I would send a z-brush v2 sculpt in for 3D printing and the result would look nothing like what I saw on my screen. This problem was cleared with mudbox and in z-brush 3 and above.)
Good luck, and I hope that helps and I hope that I am clear. I apologize to the rest of the community for rambling on. I always mean to write one sentence and then it turns into a novel 
Btw, here is a link that talks about the Dolly Shot
http://en.wikipedia.org/wiki/Dolly_zoom
anyway I do hope my anatomy is better on this one now. I did what you said about setting up camera’s for the front and side and there was a big difference from the ortho views:
