Hello Joeri,
I've figured out what is happening. In a nutshell, comparisons can handle up to 50 codes on the X axis, if more are included, the error would show up. Moreover, the data-structures that carry over the information are hard-limited to 50 codes, so even if I wrote something to avoid generating the error, the data shown would be incomplete.
This points to a possible workaround: you could move some of your codes from the root of the codeset and into an ad-hoc (container) child code. I think your root level currently contains 76 codes, so moving 26 of them into a new child code created for this purpose should work. Of course, proceeding in this way means you will have to produce multiple crosstabs, while operating on the "flatter" original codeset could have worked in a single go.
In case you need to preserve the current codeset shape, you could re-move the codes back into their original location after collecting the required crosstabs.
I do realise that this is far from ideal, but it's the only quick solution I can offer. I will discuss possible permanent resolutions with my colleagues, but I can anticipate that we won't be able to make such a solution available in a matter of days. More likely it would take a few weeks (would require to publish a new version) and perhaps more.
Please do let us know whether the proposed workaround doesn't fit your needs or if it doesn't actually work (in case I've missed another catch!). I haven't tried myself because it would require to interfere with your work/data.
With my apologies once more,
Sergio