There's a pic in my gallery (thanks if you can stick it here) showing the nut on top of the aluminium seal. I have a feeling this was how it went together. (Unfortunately I didn't get a photo as I was disassembling it 😔 )If I do the nut up tight-ish, it starts to restrict the steering movement. If I do it up just enough to stop the fork wiggling, the rubber seal underneath the aluminium disc is a bit squashed, whereas it wasn't before (see first pic). Do you think this will matter?

Thanks again JoeTBM, the bearings are 'open', if that's the right word... I have put them both balls down, as it were, because the bottom one balls up had a large gap that looked wrong, and the top one just didn't work balls up. So, both bearings: balls down, (and hopefully no balls-ups! 😂😔)
Photo of assembled headstock for your kind consideration in the album.

Well, something else doesn't look good, the plastic ring doesn't fit down onto that aluminium disc. I can't seem to find a photo of how it should go together. Any ideas?

The photo I posted above shows how the balls should be, with the open side facing away from the cups, towards the cones, so in the bike the bottom set is open side down and the top set is open side up.
This shows how everything goes together, but not very clearly:
https://cimg5.ibsrv.net/gimg/bikefor...dea1200459.jpg
It's from this Tange catalogue where it might be a little clearer.
From what I can see there's a plastic seal that goes under the adjusting race to sit against the top lip of the cup, then the race and locknut, and the other plastic seal sits inside the top of the trim piece.

Thanks for the link to the catalogue. The item is actually ZSELT. It looks like bottom one is open side up and the top one is open side down...

Hopefully I can now post a picture:
https://cimg9.ibsrv.net/gimg/bikefor...a7e88038a7.png

Per the diagram, you've got that backwards. But there's an easy test: try rotating the caged bearings on the cup or cone in one orientation and then flip the cage over. One way will cause the cage to grind along one of the mating surfaces; the other way will allow the balls to roll freely.

Or you can always assemble the headset with (upgraded; they're cheap) loose balls, thus avoiding cage-related problems. The main advantage of caged bearings is that they speed assembly of bikes in factories. The disadvantage is that there are fewer balls to distribute the stresses.

That, and the next time you disassemble the headset, unless you used plenty of grease to hold the balls in place, they'll fall out and hide in crevices all over the room.

It's difficult to tell which way up they are from the diagram, I think. I found this video helpful :
The lower bearing goes this way up:
https://cimg9.ibsrv.net/gimg/bikefor...8f03cfc353.jpg
The upper bearing is like so:
https://cimg0.ibsrv.net/gimg/bikefor...16028c243c.jpg