i read on the forum a solution is to add it from the thing channel but i can’t do this because this is some item i use in a rule and it is not directly linked to a thing channel.
Your model in general is not configured in the expected way. You have put all of your semantic points directly within a location. As you can see, this works, but only most of the time because by skipping the equipment level of the model, automatic semantic configurations are unpredictable. In general the semantic model should take the form of Location (group) contains Equipment (group) contains Points (other items). You will get a much more consistent model experience if you stick to this format.
However, as you point out you do see one place where your arrangement works and one where it doesn’t. This kind of thing happens occasionally, but without digging into error messaging more it’s hard to tell what might be going on.
Things that I suggest you check:
Look for errors in the browser console when you open the locations page or bring up the card for the location that is missing the point-item.
Check your openhab logs for any errors that might be related to the item (you can even try making a minor edit to the item, for example, adding a non-semantic tag to get the system reprocess the item which might make log error show up again).
If you upgrade to the 5.0 milestone, there is now pretty comprehensive semantic model checking that will produce informative warning when there’s a problem with a semantic item’s configuration (if you do this you will get many warnings about the improper addition of points directly to locations, but you may also get another warning about the rollershutter in question).
In the future, the best way to create semantic items will always be to go through the model page itself. Click on the location you want to add to, and then use one of the action buttons under the location details on the right side of the page. If you can’t find the error in the item you’ve created, you can delete the current item and go through this process to re-create it which will probably get it registered correctly with the model.