I tried to search for the reason of unexpected variation of CAL response vs energy noticed by Carmelo while he compared p7 reprocessed data with existing data.
During original processing we used 2 MevPerDAC calibration files:
...
All values on the plot are divided by 53.15 (actual mean ratio of MeVPerDAC coefficients of HE and LE diodes). 
This plot confirms that the use of first calibration file with lower HE/LE ratio until October 2008 should introduce 0.8 % drop of energy response at high energies. Qualitatively this is consistent with effect found by Carmelo, but has magnitude lower by factor of 2. This is still to be understood.
Average pedestal drift in HEX8 range between July and October 2008 is ~0.5 MeV and cannot explain such change in HE/LE diode ratio.