1) How to determine the ranks of each process that generated each trace
file?

Each trace file name contains the name of the machine and the pid of the
process.  If you look at the timing file, it maps ranks to machines for the
timing job that we run before and after the traced job.  We use the same
machine file for both the timing jobs and the real job.  So, we know for a
certainty that the same ranks that were on machine N in the timing jobs
will be on that same machine in the real job.  However, we don't know for
sure which pid in the real job is which of the two ranks that ran on that
machine.  This is where I think it is a safe assumption that the lower pid is
the lower rank.  

2) How do we determine the values of the parameters in the MPI calls?

Unfortunately, you can't.  Th trace of the MPI_ calls shows just the address of
the pointer arguments which (since the memory is long gone) is pretty close to
worthless.  I think it's still interesting to include the MPI_ calls bec you
can still figure out what they are doing by looking at the nested system calls.