UML22 — figure 13.12

In the latest 2.2 version of the UML spec, there was a change for issue : 11409 - redirect TimeEvent::when to TimeExpression (from ValueSpecification).
In the resolution to that issue, figure 13.13 (p427) was properly updated but it looks like figure 13.12 has a problem in that the association from TimeEvent should go to TimeExpression

At first it seems like this would be any easy resolution - just update Figure 13.12. The problem is that the Event classes in Figure 13.12, including TimeEvent, are in the CommonBehaviors::Communications package, which is merged in at L1, while TimeExpression is in CommonBehaviors::SimpleTime, which is merged in at L2. Thus, having TimeEvent associated with TimeExpression - which is actually the case in the metamodel - causes a problem in the construction of L1 (which causes issues with the generation of XMI for L1).
Now, one possibility would be to make the TimeEvent class in SimpleTime a merge increment. But the merging of typed elements has the constraint (see 7.3.40):
"Matching typed elements (e.g., Properties, Parameters) must have conforming types. For types that are classes or data types, a conforming type is either the same type or a common supertype. For all other cases, conformance means that the types must be the same."
While not entirely clear, the implication is that the resulting type is the common supertype. In this case, TimeEvent::when has type ValueSpecification in Communications and type TimeExpression, a subclass of ValueSpecification, in SimpleTime. The common superclass is thus ValueSpecification - but if you end up with TimeEvent::when having type ValueSpecification in the merged L2, then there isn't much point in typing it as TimeExpression in SimpleTime!
Another possibility would be to leave the type of TimeEvent::when as ValueSpecification, which would allow a TimeExpression to be used when SimpleTime is included at L3. But this was explicitly changed in the UML 2.2 RTF, indicating a strong desire that the type of TimeEvent::when be TimeExpression (which does make some sense).
It also doesn't seem to be a good idea to merge SimpleTime into L1 instead of L2, just to be able to have TimeExpression available for TimeEvent.
So, the proposed resolution is that TimeEvent be moved into SimpleTime. This means that time events would only be allowed at L2, not L1. But since state machines aren't included until L2 and accept event actions not until L3, it seems unlikely that this would be a real problem.