UMLR — How to model a transition to history pseudostates in two orthogonal regions?

A transition that terminates on the outside edge of a composite state is called a default entry. In this case, the default entry rule is applied e. g. the transition originating from the initial pseudostate is performed. If the most recently active substate prior to this entry shall become the active substate, the transition must be drawn to a history pseudostate. Page 564 clearly states this.

Next, the composite state is extended to an orthogonal state with two regions. Both regions contain a history pseudostate. A transition shall fork to both history states. I thought this is easy to model by introducing a fork pseudostate in the transition and drawing two outgoing transitions from the fork to the history pseudostates.

This is obviously not the case as constraints rule 3 on page 582 states: "A fork segment must always target a state. (source.oclIsKindOf(Pseudostate) and source.kind = #fork) implies (target.oclIsKindOf(State))"

It is also mentioned at other locations that an outgoing transition of a fork pseudostate must target a state. Pseudostates are not allowed as targets of outgoing transitions from a fork pseudostate.

I am confused by this rule. My question is: How do I have to model a transition ending on the history states in two orthogonal regions?

Thanks for your kind advice.