@derp_derpstein See this: Creating an isometric map - images bigger than tile
When you select ‘Collection of images’ you don’t get prompted. Instead, an empty tileset is created for you. Use the + button on the tileset window to add images.
I’m working on a project with the same challenges and I found that the best to do this currently is to use Image Objects, not tiles. The main difference is that in Image Objects the anchor point is at the center-bottom, which makes registration to the grid predictable. Furthermore, you have the option of not snapping to the grid if you don’t want that.
Because the default anchor point registration of tiles is at the center of a virtual tile of a size that is the map’s cell size on the lower-left corner of the tile being painted, this makes using images larger than the map’s cell sizes not practical because there’s no correspondence between the tile’s intended anchor point. This applies to isometric (and hex) only of course.
Perhaps as a suggestion to @bjorn: although it may be a difficult proposition, perhaps it’s possible to modify a tile’s anchor point on a per tile basis? This is similar in concept to how we can define collision shapes per-tile. Then the placement of the tile will be according to this anchor point.
And/Or as an alternative (possibly easier?) or convenience feature, we can choose a predefined anchor point, eg upper-left, center-center, center-bottom, center-right, etc.