I cannot tell if the routine is correct or not. I think you need to follow my suggestions using a selection of fixed patterns and an O-Scope to check waveforms.

The closest you can ever come to saying "the write is working" is in one of two ways. First way is to 100% verify the waveforms at the part during write cycle according to datasheet requirements including looking at all the data bits and acks to make sure they match what you thought they were. Second method (which can also be a follow-on to the first method) is to socket the part in your circuit, then perform the write patterns via your software. Remove the part and take it to a known good reading station and read out the data to confirm. One type of "known good" reading station may be a general purpose device programmer like a DataI/O unit. Another could be something such as the small parallel port adapter of PC at one time made by Philips that let a PC read and write I2C devices. The software that came with that had a provision for macroizing command sequences so that you could read/write specific registers in the attached I2C part.

Michael Karas

PS....The digital oscilloscope, in my opinion, is by far the easiest way to verify the bitbang software routines. Of course with that comes the necessity that you must fully understand the part data sheet and the I2C protocol (which is fully described in a document you can obtain from the Philips web site).
MJK