Ok, I start to understand what's going on. In this context you mustn't use
convertCoords, it doesn't do what you need: it converts a point in
target coordinates to a point in
world coordinates. Typically, it is used to know the world coordinates of a mouse click (which is initially given in pixels of the window, not in units of your 2D world).
One clue that should make you notice that there's something wrong: you had to convert leftTop, a Vector2f, to a Vector2i. When you need to perform this kind of conversion, it usually means that you badly mix coordinate systems; integer and float coordinates should never interfere.
What you want to do is the inverse: given a point in world coordinates (the topLeft corner of your rectangle entity), where will this point be drawn to the render-texture, with the given view?
So instead of calling convertCoords, you can try calling mCamera->getTransform().transformPoint(leftTop). It's a kind of hack though, this inverse conversion is not "officially" supported in the API (look at View::getTransform(), it is marked as "internal use only")
It may not be easy to understand with words, so if you don't get it take a paper and draw the world (which contains all your entities), the render-texture (which shows a small part of your world according to the current view), and ask yourself "where does my starting point belong? (world or target)", and "where does the point I want to calculate belong? (world or target)". This gives you the conversion you need to apply (world to target, or target to world), and the function you need to call.