A lot of questions seem to pop up. How do i merge multiple gradients with multiple light sources? Should the object be above or below? Below seems to be wrong with sharp gradients. On top is a bit better but the objects still would need illumination as well or it looks weird. Isnt that weird using fixed shadows?
I think the problem with the gradient (pixelated or not) is that neither is correct.
For the five "bowls", the bottom half of the image is the front and the top half of the image is the top of the object.
A light coming from behind and below the tops would illuminate neither--so you should see just an outline of the light. A light coming from behind and above should illuminate the tops (top half of the image), but not the front (bottom half of the image).
Image 1 looks more like an illuminated atmosphere--like shining a point light into a foggy area. Maybe I'm more observant of these things than the typical player would be, because I spent so much time thinking about lighting in my game. It looks neat, but the geometry of the room feels wrong.
From a geometry perspective, I think Image 2 looks a lot better. The light source is below the "bowls", so it's a believable scenario. But it's still not correct. The "bowls" should cast a shadow against such a low light. They don't, though. This only makes sense if the "bowls" are floating... but that doesn't make sense either, because the shadows they do have only make sense if they're on the ground.
I think fixing Image 1 would be easier than Image 2. Image 1 makes sense if the light source is above the "bowls" (hidden because we can't see the ceiling). The fixed shadows you currently have would work. Different materials have different properties regarding highlights, the colors of light they reflect, etc. You'd need to light each floor tile and object separately--not just apply a single gradient. But it can be done with some animation tricks and wouldn't require intensive calculations.
Casting shadows in order to fix Image 2
could require intensive calculations. The fact that I have an isometric projection in my game and you have more of an oblique projection doesn't matter. Both of us are working with 3D worlds, so casting the shadows properly would require you solve a lot of the same problems I did. The math would be a lot easier, though, since your world is completely tile-based, and mine isn't. (It's converted from tiles to walls during the game load phase). That said, it'd still be a lot of work. I don't know how much time you want to spend on it.