I would give a try to simple approach - define this process as automatic activities:
start -> import data -> filter data -> send to trajectorizer -> perform match -> calculate -> end
Every 20 minutes new process instance is create with new set of data, since all the logic resides on ESB as services I don't see a purpose of having this as never ending once process instance.
This simple process definition gives you quite good overview of what is going on and services in fact do the job.
HTH