It seems odd that the following statement is the same in both read and write functions:

i2c_SendAddress(ROM, 0);

In writeAt(), i2c_Stop() presumably initiates the self-timed write (5ms or thereabouts?). Without seeing the bigger picture, I'd ask whether you could be trying to read before the write has finished. How soon are you turning back around to read? Are you using a simple write completion timeout or are you write-busy polling elsewhere to pace EEPROM operations after the write?