Sorry to take so long to reply, sbf.
With statistics from censuses in general, there are lots of complicating factors to consider. When we see graphs, and tables of numbers, we may naturally tend think that the production of the data was as reliable, accurate and uniform as the reassuringly organised method of its display. Unfortunately there is lots of room for skewed data and misleading comparisons to arise.
Absolutely. I am only an amateur, and I am learning a lot. I have seen glossy charts and data elsewhere, and I now know how dodgy that data can be; particularly if there is a political agenda.
The format of the census may differ from year to year...
For the "religion" question, the only change (and it was a big one) was to move "no religion" from bottom to top, as shown in the following:
Extract from 2011 form:

Extract from 2016 form:

Christian groups lobbied heavily against this change before the census, but lost, because the logic of the census form elsewhere is that the most common answer should be at the top of the list.
they initiatiated a campaign to contact and encourage all inactive Mormons to identify as Mormons...
That happened in the Aust census as well. There was a Christian group that asked people to record themselves as Christian, to counterbalance the increase in Islam.
The UK census is pretty useless at counting JWs because...
Yes, I agree. I read that link. I didn't check the form but it sounds like the main fault with the UK form is that you can just select "Christian". You can't do that on the Aust form; see above. And yes, the author of that article doesn't understand JW's.
* * * * *
As a general comment from me, there were always going to be problems with the interpretation of the religion vs divorce statistics. I realised a few as I was attempting the task. Others have pointed out a few more. I think the stats are still useful, but only as a broad guide.
I plan to do a few more, when I get the time, (eg income, education levels etc) which should be more straightforward and less open to interpretation.