Long answer: In a regular old cluster there's one primary copy of each bucket that's queried by searchheads. Other searchable copies may exist, but they're not used. Using them wouldn't speed up most searches either, because the other search peers are busy serving up data from other buckets they may be primary for.
However, in a multi-site cluster search affinity lets search heads use a searchable copy in their own site, indeed speeding up searches. That's not achieved by querying one bucket in multiple search peers though, but rather by choosing one copy from a nearby search peer instead of a distant - and therefore slower - one.
Just to be clear, it never makes a single search run faster. But as martin_mueller says, it can increase capacity if you have a multi-site cluster, so the total amount of searching can go faster. It is possible that future optimizations and improvements to Splunk will allow increased capacity even without multi-site clustering, but that is not the case in the current (6.1) version.
I confirm, for now on splunk 6.0 and 6.1 even if you have multiple searchable bucket copies only one will be searched.
The reason for having search-factor>1 is to have the some buckets copies immediately ready when an indexer is lost. And not have to wait for the preparation to make one of the copies a searchable copy.