Hi! Thanks for sharing your lora. :)
If you ever feel like trying an V2 of this lora, I would suggest to avoid on the captions anything that you want to be included on the concept. For example, Rocket will always be a raccoon, have the same fur, etc, so don't describe that. The problem with your captioning is that your descriptions are so detailed, that you prompt for mcu_rckt and it outputs random things, like just now I asked for "Photo of mcu_rckt sitting on a spaceship with guardians of the galaxy uniform and an beer bottle on his paw" and it generated a feral dog for no reason xD. This happens because the training on Rocket's look have been associated to all the other words that you used to describe it.
As a rule of thumb, apart from the concept trigger word, only describe the scene, expression, style, clothing, lighting, actions... That kind of stuff, so the lora can be flexible, but better to avoid describing the look of the character so it gets associated just with the trigger word.