Your question is quite similar to one that I asked some time ago, with the additional requirement that the glue be insoluble in water, which unfortunately excludes sodium and potassium silicate. Like you, I've found Ceramabond (albeit 571, not 569, in my case) to be too magnetic. I suspect that there's an element of random variation here. It's not (as far as I know) specifically formulated to be completely non-magnetic, so there may be some batches which are more magnetic than others. Khakhalova and Feinberg (2014) investigated some other high-temperature ceramic adhesives and found that they were also too magnetic for use with weakly magnetic samples. However, they do make one useful suggestion:
It may be possible for rock and paleomagnetic labs to reduce the concentration of magnetic contamination in Omega cements, for
example, by sending the powder through a very clean Franz magnetic separator
I haven't yet heard of anyone actually doing this, but it certainly seems worth trying. They also write that
We’ve contacted Omega Engineering Inc. regarding the iron concentrations in their
cements and their technical staff are
currently exploring whether it would be
possible to produce a high purity, high
temperature cement specifically for rock
magnetic and paleomagnetic applications.
I don't know if anything has come of those attempts in the three years since the article was published, but you could email the authors and/or Omega to find out.
Khakhalova, Evgeniya and Feinberg, Josh M. (2014).
The iron that binds: The unexpectedly strong magnetism of high-temperature ceramic cements commonly used in rock magnetic experiments. IRM Quarterly, 24(2), url: http://www.irm.umn.edu/quarterly/irmq24-2.pdf