It's rumored that some items are experimental or have an initial period in which they do not count. I kind of doubt that it's true, but it's an old and widely held belief.
That said, the real exam is similar in difficulty to the NBMEs and it would not surprise me if the curve is similarly steep. On NBME 12, I looked up all the answers I was uncertain of prior to checking my score report and could not find any I missed. I missed 6 overall, but they weren't the ones I suspected. On the real deal, I know for a fact I missed 5-6 and probably had another 10-30 that I was incorrectly confident of or forgot to look up. Will find out how it turned out in 8 days, but I'm betting below my highest NBMEs.
Wow, I guess that puts my score into perspective. Wow, I feel like that is soo steep, but I guess the NBME must double (not exactly) your wrong to determine you actual score.