I recommend using strengths 0.9-0.85. Cloak is well, but other outfits are just something similar to the original outfit. Two well-referenced images in the dataset are not enough for me to make it close to the original. Good photorealistic images are hard to get by photorealistic checkpoints and/or LORAs, but semi-realistic or 3D and especially 2D work much better. If you're going to use another LORA style, I recommend increasing the strength of it more than you usually put, but keep the strengths of that LORA no lower than 0.8.