Unity UI extensions mouse drawing: annoying offset between touch position and drawing position

I think I found a solution to my problem or, better, an approach to follow when you want to design a 2D game where a Canvas covers completely the phone screen and the camera is rendering exactly the phone screen. With this approach I got rid of many problems I was facing with coordinates conversion.

Here are my advices:

1) Design you interface (buttons and the like) using a screen size (in Game tab) of 1920x1080 or 1080x1920 (landscape or portrait)

2) Set you camera in orthographic mode
with a size of 9.6 (100 pixel per world unit)

3) Add a Canvas to hold your UI
elements with Render mode set to Screen Space - Camera

4) Add a Canvas Scaler script and set a reference resolution of 1080 x 1920 (or opposite for landscape) and set the Screen Match Mode to Expand

With these settings everything is coherent if you use a phone with a resolution of 1080 x 2340 but if you try with another phone resolution (for example 1080x2340) conversion of coordinates will start to become weird. Then I did the following

5) I setup the camera in the Awake method in a script associated with an object of the scene (can be the camera itself) with the following code:

 void Awake ()
 {
     ...
     Camera.main.orthographic = true;
     if (portrait == true)
     Camera.main.orthographicSize = (Screen.height*1080/Screen.width)/200.0f;
     else
     Camera.main.orthographicSize = (Screen.height*1920/Screen.width)/200.0f;

     ...
 }

6) I set up the canvas in the Start method of a script associated with a game object in the scene with the following code:

 void Start() {
     ...
    canvas.GetComponent<CanvasScaler>().referenceResolution = new Vector2(1080, Screen.height*1080/Screen.width);
     ...
 }

7) Because I have a background image that has a resolution of 1920x1080 I use also to change its local scale with the following code:

     float cameraHeight = Camera.main.orthographicSize * 2;
     Vector2 cameraSize = new Vector2(Camera.main.aspect * cameraHeight, cameraHeight);
     Vector3 extents = background.GetComponent<SpriteRenderer>().bounds.extents;
     background.transform.localScale = new Vector3 (cameraSize.x/(2*extents.x),  cameraSize.y/(2*extents.y), 1.0f);

With the previous steps (valid for portrait) now I can draw with my UILineRender and using the ScreenToWorld method with no offset at all.

I hope this can help.

Antonio.