Thank you John and Martin for answering my plea about posting links on the forum.
I wonder if people's different experiences are down to timing? Dies wear out, as do moulds for plastic injection, and probably not at the same rate. The Exactoscale chairs and Hi-Ni rail I bought six or seven years ago worked well together - there was a bit of curvature on the base of the chair, easily resolved when fixing in place with Butanone. I cannot say the same for the chairs and rail which I bought in September 2017, which were the subject of the trials described above. Has anybody else got experience of using Exactoscale chairs with Hi-Ni rail, both of them bought within - say - the last 18 months? Did you have similar problems with distortion and to what extent: as bad as sample 1, 2 ,3 or 4?
When I looked again at Sample 1 through a jeweller's loupe under a good light, I could see daylight through the chair's jaws - the only points of contact were round the foot of the rail. We know from the drawing above that the foot of Hi-Ni rail is sound, so to my mind this confirms that the chairs are part of the problem (or at least this batch is). Maybe things will improve now that Andrew has taken back control of Exactoscale products.
On a more upbeat note, I have now fixed each sample to a ply timber with Butanone, and I have to say they have all come out well - bases nice and flat and rail correctly inclined. I will try a length of rail threaded with chairs straight from the sprue, although I am still a little concerned about how this will work during the gluing process - when some chairs will have been glued flat and others - just a few centimetres away, will still be bent like a boomerang.
David is quite right to ask if all this is actually a problem at all. I would say it is: even if things turn out right in the end, surely P4 modellers (and even EM modellers, like me!) shouldn't have to rely on forcing things into the right shape with a firm hand and a powerful solvent?