Regular expressions in pattern matching can accomplish this despite the varying number of characters --- regular expressions are pretty flexible! I use online tools like https://regexr.com/ to experiment since it include explanations.
Note: settings in any given regular expression tool may differ from workflow defaults)

Since we want to replace the last dash - or underscore _, I created two patterns: one to capture the part before the final dash/underscore while also excluding the numbers and _ at the beginning, and a second pattern to capture the section after the final - or _

You could then combine the two tokens using something like this:

%(PatternMatching_BeforeSlash)//%(PatternMatching_AfterSlash)

Where the // inserts a single / , since forward slashes must be escaped using another / in workflow token syntax.

The following pattern will return everything between the first underscore _ and the last dash -

(?:[^_]*_)(.*)(?:[-_])

(?:[^_]*_)

• ?: specifies a "non capturing group", meaning things that will be excluded from the final match
• ^ inside [ ] means "not the following", so [^_] will return characters except _
• the first non-capturing group is then "everything up to and including the first _"
  • You said in your post "from the right" but your examples were from the left, so that's what I based this on

(.*) captures the rest of the text up until the final group, (?:[-_]), which is a non-capturing group for the last dash or underscore. (Again, I based this off of the examples, which contradicted the description). (.*) will include prior - and _ because .* is a "greedy" match.

The following pattern will return everything after the last dash or underscore

([^-_]+)$

The $ indicates end of the string (or end of the line, if multiple lines are involved. In your case, for folder names, they should be single lines).

[^-]+ uses the plus + to indicate "one or more". Combined with the $, it will find all characters until it hits a - or _, but starting from the end of the line.