Problem with RFDF matrix?

All, 

I would like to model a thin-lens deflecting mode cavity. If I set L=0 in the
RFDF element and inspect the matrix provided by runin the testrfdf.ele command
file, I get the usual think-lens matrix for a deflecting mode cavity (TM110). 

I inserted this matrix into a beamline and ran comprfdrMAT.ele, I also did
tracking using the RFDF element (see comprfdf.ele) the results are different and
I cannot undersdtand why; see runTEST script. 

I saw there were issue with a previous version of elegant and I am using elegant
V24.1 (darwin-x86 version of Oct. 6, 2011). 

Thanks for any help. -- Philippe. 
