See, this is exactly the type of thought that I just wasn't having, thank you Tim! I really hadn't considered that approach, ultimately though probably not being familiar enough with word lengths that aren't 8-bit.
Initial results from a quick table dumped out of excel appear to have reversed the behavior though,with DAC-A behaving, and DAC-B showing a corrupt (though much less so than previously) sine.
With this in mind, I'm going to double check my bitMath in excel, and revise the LUT in the morning, after a good nights sleep.
Thanks again Tim, I'll let you know how I get on.
Cheers,
Tom