Here's a question I've been getting a lot recently. I've added it to the FAQ.
When you have multiple images in one locus, how do you remember their order?
Since I use "single-image" systems -- each digit/card/etc group always translates to a single image -- I must still recall the order of images within each locus. Contrast that with PA- or PAO-type systems, which by definition give you the sequence.
I deal with this mainly by focusing on story directionality. For example, say I'm looking at 375 - 004: Michael Jordan - scissors. I might imagine Michael Jordan using scissors to cut the locus. 004 - 375? A large pair of scissors are cutting Michael Jordan as he lies on the locus.
For some image pairs, that process doesn't work as cleanly. In that case, I focus on making the images interact directionally in space -- left to right or top to bottom. Take 536 - 609: lime juice - Empire State Building. I'd imagine a lime juice container hovering above and squirting sour lime juice down on the ESB. 609 - 536? Perhaps the ESB falls to its right and squashes the lime juice container. Or maybe it's dropped from above and crushes it.
Granted, I still occasionally make "swapping" mistakes, but following these is usually enough to ensure I recall the correct order.