From my initial testing (pics should be up in an hour) it seems to apply the pattern to anything prefixed with "dinosaur ", but also seems to be also applying the base model's default understanding of dinosaur prints on top of that (i.e, often wind up with "normal" dinosaur print alongside what this makes). I've seen this ability to apply patterns to other things occur with clothing LoRAs before, and I think it would be worth considering working with as a feature in the future (e.g., use a non-contaminating trigger trained as prefix, multiple different articles of clothing with same pattern in training data)