I'm working with nested XML input where elements like <title> contain a language attribute, e.g.:

<myElement>
  <title language="fi">Automaalari</title>
  <title language="sv">Billackerare</title>
</myElement>

In the past, I successfully used this mapping to generate properly language-tagged literals:

rr:predicate dc:title ;
rr:objectMap [
  rml:reference "./*[local-name()='title']" ;
  rr:termType rr:Literal ;
  rml:languageMap [
    rr:class rdf:langString ;
    rml:reference "./@*[local-name()='language']"
  ]
] .

However, in the latest version of RMLMapper that I tried (currently using the cli jar), all language literals that are using this kind of mapping are missing from the output.

I have tried to modify the mapping based on the test case
as follows:

rr:predicate dc:title ;
rr:objectMap [
  rml:reference "./title" ;
  rr:termType rr:Literal ;
  rml:languageMap [
    rr:class rdf:langString ;
    rml:reference "./title/@language"
  ]
] .

but all language values appear to be taken from the first <title>, resulting in:
dc:title "Automaalari"@fi , "Billackerare"@fi .

This seems to indicate that rml:languageMap does not correctly align language attributes with their respective values when dealing with sibling elements in XML.

I've tried adjusting the reference paths and iterators, but results didn't change. Either the attribute was missing completely or all attributes were getting the same language tag.

Is this a known regression or change in how rml:languageMap is evaluated with sibling XML elements?
What is the recommended way to generate correct language-tagged literals from repeated XML child elements like <title> in the current version of RMLMapper?