The problem is normally the combination of the data-compression applied to the analogue waveform and the shift in the timebase that occurs when the resulting data gets turned into a packet transmission. Modems and faxes have fairly wide bandwidth requirements and expect constant transmission delays without any jitter. VoIP calls don't normally provide either of those. I've tried faxing over VoIP and even with the widest bandwidth codecs it was rarely successful.My impression is that modem tone does not seem to be supported on any voip connection. I suspect the way it compresses all the quiet bits in the call etc may be the problem.
There's a good article on modem/fax over VoIP at http://www.soft-switch.org/foip.html