This is a long one, but I hope this can be a comprehensive explanation of what we want and why it's absurd that we have to ask for it.
To ensure there are no misunderstandings, "Just to clarify, it is not that we want to place decals without scaling them" could be misleading. We do want to be able to place decals and have them start at the same scale and size as their image so as not to have to scale them. But we also want to be able to scale them accurately.
When you manipulate and export an image, its size is embedded into it, commonly in the form of resolution in pixel width and height plus the pixel density in the form of pixels per inch or per centimeter. Also, despite being infinitely scalable, SVGs are usually made in a workspace with set dimensions and are exported at their specific size. When imported, the user expects them to start at that size, unless the user otherwise specified somehow, such as something like a "fit" option.
User interfaces should be analogous to other familiar and similar interfaces. These are the things that the general user expects it to be analogous to for minimal frustration, as well as the things it needs to have to be useful. (The following also should apply to the canvas and, to some extent, the import SVG features)
Some of these ways I'm going to go over you already have covered, but there are some expected functions here that you are missing, which would be nice to have, and some that you need to have. You should take note of them.
The Sketch interface.




It's assumed that it would behave very similarly to the way basic interactions with a rectangle in sketches work, albeit with a somewhat simplified workflow. It's expected that, like a rectangle, you can freely stretch it from its edges or dimension its sides to a specific known value number. With rectangles, you can constrain its edges, middle of its edges, corners, or midpoint to other geometry or have an offset to those points. When you move any of its points on top of other geometry, it automatically applies a coincident constraint. Or, if you remove the horizontal/vertical constraint, you can rotate it from any point freely, or dimension it to a specific angle (sometimes this requires you to constrain the point you want to rotate it about). It's a parametric modeling app; it's expected that you can use hard parameters to define it.
I think users expect also expect that you at least could draw a rectangle in a sketch, and then when you place the decal, you are given the option to have it automatically fit to the rectangle (or face) you click on. If you wanted to go the extra mile, also give an option for the rectangle (or face) to instead act like a bounding box that crops the image non-destructively similar to how photo editing apps work, that would be awesome.
However, the decal is also an image; it's expected that you can interact with it in the basic ways that you can in nearly every software that lets you place and edit a photo.




Adobe Illustrator:
This, and other similar apps, are where many people will be coming from with their custom decals. You can align and snap the photo to other existing objects by their edges, center, and corners (just like sketches!.... but not in decals....). You can grab an edge and rotate or scale it freely (Just like sketches... but you can only do it from the center in decals). You can input specific widths or lengths individually like you would dimensioning it in sketches (like you can in sketches... but not in decals...). You can set a specific angle via a text field, with options to rotate from the corners, center edges, or center (like you can do via sketches if you constrain things right. You can only do this around the center in decals). Also, you can lock and unlock the width/height fields to change whether one dimension scales proportionally to the other without doing any of the math yourself (not in sketches per se, but a crucial photo editing feature). You can flip the image horizontally/vertically with a button (You have this in decals, nice!). You can control the opacity (You have this, nice!). You can specify a coordinate Y/X distance based on the origin (this could be based on the center/origin of whatever face/plane you selected to place the decal on).
Let's look at another example that you aren't going to design the image in; only place it: Google Slides.


Yes, there is more expectation of visual design here, so we see other options like recolor, brightness/contrast, drop-shadow, etc. However, Fusion has the same expectation of layout control, so we see that Google Slides lets you do the same as Illustrator as mentioned above; aside from that, it is missing the ability to rotate from anything other than the center, and once rotated, corners only snap to edges, and not other corners. A lot of the same goes for Powerpoint, Photoshop, Gimp, DaVinci Resolve, and even MS Paint, etc. Every basic built-in photo editing app on every device pretty much at least has the ability to control the images like that "Size & Rotation" panel.
To illustrate how intrinsic the ability to grab a photo's edges and change it to a known size is, even this forum interface lets you do so, albeit not with a way to type the number in. The same goes for many email clients and word processors. It is beyond absurd that you cannot do this.

I cannot stress enough how awkward and less useful it is that it is not analogous to either of those things, especially regarding alignment relative to existing objects and setting to a specific, known size instead of only a relative one. It is incredibly jarring and weird that it is not because this is the expected behavior of every app that deals with photo editing and every app that lets you do any form of parametric modeling.
If we design a decal/canvas at a specific size and/or want it to go into a specific space aligned to a specific thing... We can't... We just have to eyeball it. Why?
Regardless of how important or unimportant the Fusion development team thinks accuracy dimensions and placement is with this feature is (accuracy and geometric/numeric definability in placement is literally why people model parametrically, it's what makes it parametric), I think it's clear to say this feature is an utter failure in its user interface.
Also, keep the scaling percentages based on the size the image imports in, the way it works now always making it so that each scale and especially distance set as the new 1.00 scale and 0.00 [unit] distance an after you confirm the choice and come back later drives us up the wall. I know it's clearly a symptom of the fact that you made this feature arbitrary and not parametric, but I just really need to hammer in this just in case that aspect would otherwise somehow be preserved.
I worry that there might be some sort of survivorship bias if you're looking to see people how often they use the decal feature and what workflows they use it in. I think you'd find that people would use it a lot more if it didn't suck.