. . . which feature(s) takes precedence?
Human morphology undergoes dramatic change in the history of an individual. Comparative morphology, if it is dealing with specimens of the same age, should reveal consistencies in bone structure, but these consistencies need be no more rigid than comparisons between Shaq and a dwarf. I think number and shape of bones is a good rule of thumb. What else do we have to work with after decay has had its way?
Fester, you are asking some good questions.
The first goal of a morphological study in osteology, and probably other fields as well, is to establish the regularities of your group--figure out the traits which characterize one group while at the same time excluding other groups. There are several good statistics to help with this; the one I am most familiar with is multiple discriminant function analysis. Its goals are to group things, and to figure out if any (or which) traits help to do this. It is used in evolution and population studies quite a bit. The Bantu expansion in Africa was tracked using this kind of osteometric data and analysis.
Shape generally takes precedence over size! There are even ways to find shape similarities between male and female (ignoring size and gender differences) while excluding males and females from other groups (relying again on shape differences).
Once the regularities of a group are established, with the characteristics that define that group understood, it is easier to work with the outliers. Often size and gender differences are not very critical, when compared to shape differences.
This type of study is generally grouped under osteometrics--measurements of bones. There is another whole field which tracks genetic markers in bones. As you might expect, it has its own multivariate statistics to help out. I have not actually done any of this (I prefer measurements).
A lot of this is becoming obsolete with the genetic studies.
To summarize, a lot of what osteologists do is computerized, which takes a lot of the subjectiveness out of the process. Regularities are more important at first than outliers, but after the base population is understood outliers are much more fun. These things are serious studies, with thousands of scientists and aspiring-scientists working their tails off to figure it all out. Don't try this at home!
Hope this helps.