DNA Base-Flipping: TI vs PMF

I would like to calculate the free energy change for the process of base-flipping in DNA using Amber. I have been working with the online tutorial of thermodynamic integration in Amber and have a few questions as to how I could implement this for the base-flipping of damaged DNA. Along the lines of the tutorial I was thinking of a cycle in which DNA (flipped-in)-->water and DNA (flipped-out)--> water using dummy atoms. Can Amber handle using that many dummy atoms (DNA = duplex 16-mer)? How could I ensure that the two "water" states are the same, or can I? To get around the waters being the same I was also thinking of starting with a water box that contained both the flipped-in and flipped-out DNA (as dummy atoms) and then going both the flipped-in and flipped-out directions to get the free energy difference. The larger question is - How do you determine that the reference state (water) is the same for any cycle? Also, could this be done directly (without a common reference state)? If yes, can Amber do this? I am interested in using thermodynamic integration rather than potential of mean force for this calculation because of the choice of path for the flipping process. Any thoughts?? 
 Lauren L O'Neil
 lhunter2 (a) nd.edu