Sure, makes sense. Let me explain just a bit what I meant in the 2nd PP…
In the shot in post #3, the inside of the window-frame is here. In post #1, it is conspicuous by its absence. But the only significant source of light for that frame probably comes from the outside glare. (That’s your decision of course.) Therefore it needs to be part of the shot in post #1.
I’m suggesting that you set up the shot in two separate renders, generating each one to a MultiLayer OpenEXR file output. The first is post #1. The second is post #3, but with a hole where the window (and frame) will be, and no light spilling on the floor. (But post #1 does contain a reflection-only plane coinciding exactly where the floor is.) You now perfect the shot in post #3, getting the shadows in the corners just right.
Now, you create a blend-file, with the node-based renderer and an Image input-node pointing to each one of these intermediate files: the splash of light (and window), and the room. The file-format I suggested is specifically engineered to capture all of the digital data, separated into render-layers … it’s a digital data format. You can now blend the two images together, perhaps just with a Mix node but maybe with other color-correction steps; the world is your oyster.
You simplified your workflow into three steps, corresponding to three concerns, so that you can deal with each one separately: a clean halo of sunshine (and the window); a room into which the sunshine will flow; and mix-down of the two to produce the final image (or images).
This is “the ultimate digital darkroom,” and an eye-opening critical point to recognize … how the magic-tricks actually work. In the world of CG as in conventional photography, “the image might be captured in the camera, but it is made in the darkroom.” When you take a render and “it doesn’t look quite right to you,” realize that what the renderer produces is really a component, usually one of several, waiting for that final “darkroom” step. Computer graphics uses the same kind of “clean recordings followed by mix-down” process that’s used to make songs.