Your pairwise divergence values already constitute a distance matrix. If you just have a list of values, you'd have to sort them first into a tabular format, that's all that is required to have a matrix. (Depending on downstream processing, the exact format you want to get at might be different though.)
One thing to consider is whether you might want to do any data transform between raw values and analysis.
Then, for "phylogenetic" analysis (I presume you want some sort of a tree?), you could use any of the common clustering methods such as UPGMA or NJ. They are implemented in all sorts of software (e.g., PAUP*, PHYLIP, MEGA, FAMD, NTSys-pc, .. and there's bound to be an R implementation). Also, many people might use principal coordinate analysis (PCoA).
If you have data compatible with Phylip suite of programs for distance calculation you may use various methods offered by Phylip. For example I have used kitsch program in a similar context for generating distance matrix.
From Phylip Wikipedia page: fitch : Fitch-Margoliash distance matrix method with molecular clock. Estimates phylogenies from distance matrix data under the "ultrametric" model which is the same as the additive tree model except that an evolutionary clock is assumed.
You may also take a look at variety of distance calculation methods offered by phylip here.
EDIT: Soon after I posted this answer Josh Patterson pointed me to a presentation (.ppt file, 45MB) that explains a variety of distance measurement methods that you may consider for your calculations. The original tweet by @jpatanaooga is here.