Why do some journal editors request that we present the main effects when the interaction is also significant? Is "for sake of completeness" a good enough reason?
Well, the reason you ran an ANOVA in the first place is, presumably, because your hypothesis was predicated on the finding (or not finding) of main effects. Yes, of course you would report interactions, although with respect to that, you may want to consider an ANCOVA or MANCOVA. Remember also, that with the GLM, regressions and partial correlations may reveal things in a more succinct way, and especially the latter can be useful when attempting to partial out effects.