Mid journey is already fairly heavy handed in terms of its onerous restrictions and puritanical censorship (blood, gore, etc), so this comes as no surprise.
I find myself having to turn more and more to stable diffusion models to finish a lot of my concept work.
In my opinion MJ v5 has better generations on average than a well-customised SD pipeline, but only by a little bit. SD definitely requires more work in terms of set-up and greater specificity in your prompt text.
I would recommend downloading one of the more aesthetically pleasing models from Huggingface or Civit.ai, such as "Deliberate", "Protogen" or "Realistic Vision".
Using 'clip interrogator' on existing images that you find appealing can be a great way to find interesting keywords to include in your prompts.
If you're comfortable hooking things together in a node based UI, I would recommend ComfyUI as it can give you greater flexibility around how your images are generated.
What kind of images are you trying to generate? I can maybe give some more specific advice on that.
I find myself having to turn more and more to stable diffusion models to finish a lot of my concept work.