I think perhaps you're putting too much into the "Map Reduce" term. It probably doesn't mean the same thing as other places where it you've seen it used. In the Splunk world, it means search queries are sent to all indexers for processing. The indexers send their results back to the originating search head which combines the results and finishes processing the query.
I already tried going through that and the other whitepaper "Technical Paper: Large-Scale, Unstructured Data Retrieval and Analysis Using Splunk".
However, I am still confused on the overall utilization of Map Reduce in Splunk.