These graphs are not based on two data partitions, but one (B, only intergenic spacers, ie. noncoding parts) is a subset of the other (complete plastome, A). Wouldn't it have made more sense to compare an amino-acid/codon-based tree vs. one based only on the non-coding regions?

This may have helped regarding "advantages of using chloroplast genome data to illuminate those phylogenies that have not been well solved yet by traditional approaches in other plants". In which aspect does the new complete plastome data resolves intergeneric relationship further than earlier analyses? In our cited 2013 paper (Grimm & Renner, 2013), we showed that plastid data confirms the sister relationship between Alnus and Betula inferred by an earlier study using high-variable nuclear markers with (comparably) high support (the cited Forest et al. 2005). Both studies placed Ostryopsis as sister to Carpinus-Ostrya (your tree in A). Only the study by Li et al. (2004), using a single placeholder per genus, and aiming for a all Fagales tree, failed to do so because they included wrong sequences for Alnus.

Since the complete plastome tree (A) agrees with earlier morphology-, nuclear-, and plastid oligone-based hypothesis, the alternative seen in B may be a branch attraction artefact due to the use of too divergent data, less-fitting model.