Two ideas come to mind for how a Grignard reagent could differ from an organolithium compound. Both are related to the bulkier nature of the Grignard molecule.
- Magnesium forms a more ionic bond than lithium.
In this hypothesis, the reason the organolithium compound attacks the carboxylate is that we have significant covalent bonding between one oxygen and the lithium, so we don't really have an ion: the carbonyl function is preserved with the non-lithiated oxygen. If the $\ce{MgX^+}$ moiety from the Grignard reagent were bonded more ionically, then the Grignard substrate would behave more like a true carboxylate ion with poor electrophilicity. The larger size of the Grignard fragment (including the halogen atom) combined with only one net positive charge (rather than two as on a bare magnesium ion) could promote such greater ionic character.
The diagram below shows how the extra atom could promote more ionic character in the Grignard product (using bromine as the halogen). Assuming magnesium bonds covalently primarily with its $3s$ orbital, that orbital forms a three-center four-electron bond with the more electronegative neighbors, which gives an ionic contribution to each individual linkage. With the corresponding lithium compound a similar delocalization does not occur because the lithium is bonded only to a single atom.
Note that a similar delocalized bonding model can be rendered for the Grignard reagent itself, thus enhancing the ionic character of the Grignard reagent's carbon-magnesium bond.
- The ketone precursor with a Grignard reagent is sterically hindered.
When an organolithium compound reacts, two $\ce{OLi}$ groups become attached to the original carboxylate carbon and these are later hydrolyzed to a carbonyl oxygen. With a Grignard reagent, two bulkier $\ce{OMgX}$ groups would have to be attached to form the ketone precursor, creating more possibility for unfavorable steric conditions.
