I wasn't able to figure out a perfect solution, but I managed to simplify your regex a bit and to exclude the unwanted parts from tagging by removing the space (\s) at the end of the regex:
^[A-Z]+\d*[A-Z]*\d*(?:\*\*)?

From what I can see in the preview, it isn't about the greediness per se since it creates actually two tags, the first one wanted and the second one unwanted, but what Studio considers as the start (^) of the string. My theory is that once Studio has tagged a bit of text, it continues parsing the remainder of the string and considers the position after the last match as the new start of the string. That would also explain why adding a negative lookbehind at the beginning of the search pattern (?<!\s) doesn't have any effect.

I hope this helps a bit!

Greetings from Brussels,
Raphaël

Raphaël Toussaint

Thanks for your effort. I've tried what you suggested, it doesn't seem to work.

Is anybody of the SDL team here to help me out with that?

Thanks

Strange, as you can see from my screenshot, it worked for me (Studio 2017)...

Maybe Paul would be able to help.

Try ^(?n)([A-Z]+\d*[A-Z]*\s?)+
A comment concerning style. (\d+)? (=one or more digits, but this is optional) is more or less equivalent to \d*
Although Regex101 is a nice testing tool, it is .NET-conform, the regex flavour used in Studio (RegexHero (free) and RegexBuddy (licensed) are two alternatives).

Try ^(?n)([A-Z]+\d*[A-Z]*\s?)+
A comment concerning style. (\d+)? (=one or more digits, but this is optional) is more or less equivalent to \d*
Although Regex101 is a nice testing tool, it is .NET-conform, the regex flavour used in Studio (RegexHero (free) and RegexBuddy (licensed) are two alternatives).