[CobaltBlue]

When you use the words "more fit" you use them in relation to a point of stability.

No, as I understand mutations there is never stability, per se. The target is always moving.

If the "best" jaw length is 1 meter, then everything will be driven towards 1 meter from every direction.

I have no idea what the above sentence means. What does "best jaw length" mean? What does "everything will be driven" mean?

In evolution, as in everything else, there are trade-offs.

Jaws need muscles to open them and close them. Jaw muscles need to be anchored to the skull. There needs to be a tongue proportionate to the jaw, and the tongue is also anchored to the skull. There are sinus cavities, and nerves and blood vessels and skin, and countless other bits and pieces. Teeth. Tooth sockets. A random mutation in the length of the jaw bone doesn't carry everything else along with it.

Lengthening the jaw will probably increase the ability to shred meat but decrease the ability to chew, by throwing the muscles out of whack, so the ability to utilize food sources will change.

Lengthening the jaw without lengthening the tongue will make it harder to swallow. Lengthening the jaw without lengthening the sinus cavities will make it harder to breathe.

[AndrewC]

No, as I understand mutations there is never stability, per se.

Gee, are you now arguing on my side? Read what I have written. That little realization then makes one wonder why anyone could think that something complex could ever evolve. Simulation programs that "prove" evolution always have a non-moving target as the goal. They never have a varying goal. For instance the target chosen by one is "Methinks it is like a weasel" which never changes, not:
methinks it is like a aardvark,
weasels are like otters,
methinks it is like a vessel,
wethinks tis like a weasel,
as the goal at various times.

Having a varying goal results in chaos for these programs.

[Stultis]

I have no idea what the above sentence means. What does "best jaw length" mean? What does "everything will be driven" mean?

AndrewC is using some non-standard terminology, but seems to be talking about a standard idea, originally introduced by Sewall Wright, of an "adaptive lanscape". This is a topological schemata where the Z-axis represents relative fitness, and the X & Y axes (sp?) represent genetic or phenotypic variation. In a given environment, X/Y coordinates of high fitness represent "peaks," and ones of low fitness represent "valleys". AndrewC causes confusion by writing about "stable" environments, but of course virtually no real environments are "stable" in the sense that peaks and valleys in the adaptive landscape remain fixed.

A random mutation in the length of the jaw bone doesn't carry everything else along with it.

Actually it usually does, in fact, do exactly that. This is due to mechanisms of embryological development in which, for instance, architecture develops according to position along chemical gradients, and the growth of structure is influenced by, and coordinated wrt, adjacent structures.