As far as I can tell, the 50 figure is somewhat erroneous/only approximate which is (understandably) caused by the confusing way the proposed landing site list changed during the selection process.
I've attempted to collate the information from the Landing Site Workshops overview which has all the presentations and announcements from the entire process. In particular the First Workshop site list, the Second Workshop Program and the Workshop Announcement which states (emphasis mine):
The primary goal of the second workshop will be to evaluate the 33 sites that emerged from the first workshop as well as any new sites proposed within the framework provided by new Mars Reconnaissance Orbiter (MRO) data for the sites, the science requirements of the MSL mission, and a better understanding of the MSL engineering requirements and the safety of the landing sites relative to these requirements.
Before the start of the 2nd workshop, 3 more sites had been proposed, increasing the total to 36 sites. By the start of the 2nd workshop, 9 sites had been dropped and a further 24 sites had been added or sub-divided from existing sites. This gives us a total of 60 sites considered with 51 sites lasting to be discussed at the 2nd workshop.
These changes are mostly outlined in the General Assessment of Safety of Prospective MSL Landing Sites.
In any case, by the end of the 2nd workshop, the shortlist of 6 sites had been selected. It's interesting to note that all the shortlisted candidates were proposed in the 1st workshop.
Collated (raw data on Dropbox) with lots of help from @BlueCoder:
Note: this table is my own work from summarising the publicly-available data on the workshop site and therefore may be incomplete. Thank you to @BlueCoder for fixing all my mistakes! Please use the raw data freely