OK, so what I've assumed were crop marks (as in, targets for cutting, so should lie within the bleed area) aren't at all - they always line up with the edge of the document. I just printed out a pdf and measured them. So what does setting up a document with a (eg) 3mm bleed achieve, if it isn't represented in the final pdf? I still had to manually 'oversize' my document, and may as well have not set up the bleed in the first place. Slightly confused.
Walt, in Test1 I overlaid the image on a gradient, in a doc set up as 65x240 with 3mm bleed. those trim marks I thought were crop (bleed) marks, so I was confused that they didn't intersect the gradient. So, I did test3 - which is a gradient larger than the artboard, but nothing else. the trim marks now intersect the gradient as I had expected. When I took that doc and added an image, the pdf output was constrained within the trim marks again. I haven't included an image of that, because frankly by that point I'd lost the plot.
Anyway, I've worked around it, but am left feeling that the bleed issue may as well be completely ignored if we can't see it on proofs or final artwork, and have to manually oversize to take it into account. Ahem. i'm off to look out my Crayola set...
Test3.pdf
Test1.pdf