hitting rate_rule with rank4 when training reaction valid

I encountered this potential issue when deciding precedence of rate rules and training reactions.

This reaction C=C + C=C[CH2] -> C=CCC[CH2] hits node [Cds-HH_Cds-HH;CsJ-CdHH] in R_Addition_MultipleBond. For this reaction & node, there is a training reaction for it, and two rate rules. If none of the atoms are isotopically labeled, it obtains the training reaction. If an atom is labeled, then it gets the rate rule (which is all as expected).

The rate rule with higher rank of 4 has the comment "Aaron Vandeputte GAVs CBS-QB3", which leads me to think it should be less trusted than the training reaction. However, the algorithm uses the rank 4 rate rule instead of the training reaction, which does not seem like the ideal kinetics choice.

Reproduce the precedence case for C=C + C=C[CH2] -> C=CCC[CH2] at node [Cds-HH_Cds-HH;CsJ-CdHH], comparing unlabeled and isotopically labeled atoms. Trace how the training reaction and rank-4 rate rule are selected; done means the intended kinetics choice is established and covered by a regression test.