The model does not correct mistakes when i want to refine an image it produced

I asked the model to create a dress using a certain fabric, which i uploaded a reference image of. The model made the background the same colour as the pattern of the fabric. The background should have been plain. In refining the image, when i asked the model to correct the background and make it white, it left it the same as before. When i asked the model to remove the scarf around the model's neck, it left it there. These are big bugs for a service which is supposed to be the best in the industry.

Also, am i allowed only ONE reference image? What if i need the generated image to reference 2 or more images? With these limitations, why would i pay for Firefly? I think the complementary generations should allow the (almost) full functionality of Firefly so that you can make an educated decision on whether you want to pay for it.