In the light of the last revelations, this may or may not be. It will depend on how the cells are actually written to at the low, hardware level. How big is the writing granularity? If I overwrite a 0xFF with a 0x7F, is just one bit actually flipped in hardware, or is a new writing pass performed over the remaining seven 1s, regardless of whether they were already 1s to begin with? However, if overwriting a 1 with a 1 doesn't count as EEPROM wear, then 0xFF or 0x7F would perform equally well.