5.9 years ago
Yes, this is possible. It is one of BEAST's strengths.
The key is to fix the starting tree so that the node you wish to place an informed prior age on exists. For example, you would specify a split between the outgroup and all other ingroup taxa; the age of this node would be 80 million years. You can either fix the age of this node or, more reasonably, specify a prior distribution of ages. If you know that 80 million years is a mean estimate of the age of the split, for example, you might place a lognormal, normal, or uniform prior on the age of the node. The posterior distribution of trees will be estimated under these constraints (because you are, after all, doing proper Bayesian inference by using an informed prior).
However, with only a single calibration, your 95% HPDs on other node ages are likely to be somewhat large. How large directly depends on how much sequence data you have. If you have too much data for a Bayesian analysis, I recommend estimating a tree using RAxML and then transforming the branches by placing prior ages on your node(s) of interest using TreePL.
There are many excellent tutorials to do everything I have described. Search around on Google a bit. Good luck!