get output by label and input returns null for nodes that are not in an ancestor-descendant-relationship

I'm having a problem with the generator and I suspect it may have something to do with the order of the micro-steps which I don't fully understand, although I read the User Guide section about the generator algorithm.

I have a concept, let's call it ConceptA and its reduction rule looks something like this:

concept ConceptA --> <T  callA($LABEL$ MyLabel [someCall($COPY_SRC$[condition])], $COPY_SRC$[exprB])  T>
inheritors false 
condition <always> 

As you can see, a mapping label is attached that maps the instance of ConceptA to some output.

Now the problem is the following: The second COPY_SRC, the one which has the dummy text exprB under it, is not under the label macro and copies some input src to its place. This input is of another concept, let's call it ConceptB. ConceptB has of course its own reduction rule, but also a weaving rule. The purpose of this weaving rule is to weave additional arguments into the call expression whose dummy function name is someCall in the example above (so the extra arguments should go next to the node labeled condition in the example above). In this case, the returned sequence of the foreach of the weave-each of the weaving rule is an empty sequence, so no additional arguments should be weaved in in this particular case. But the query to retrieve the context for the weaving rule is executed nonetheless, which should be fine. The code for this query looks essentially something like this:

genContext.get output MyLabel for (node.ancestor<concept = ConceptA>)

But this returns null for some reason. And so I get an exception:

java.lang.IllegalStateException: @NotNull method jetbrains/mps/generator/impl/interpreted/TemplateWeavingRuleInterpreted.getContextNode must not return null

I expected this to work and not return null. I expected that the reduction rule for the ConceptA instance is applied including the attachment of the mapping label, and only in the next micro-step are all the rules for the subnodes executed that got inserted via COPY_SRC. But now that I get this exception, I wonder if this assumption is wrong. Maybe the weaving rule for the ConceptB instance and the label attachment are executed in parallel since they are not in an ancestor-descendant-relationship in the AST. Can somebody tell me if this should work or if the execution is indeed parallelized and therefore the search for the output by mapping label is unreliable?