The answer depends on your definition of efficiency. The advantage of the system with the potential closer to the thermodynamic potential (-0.35 V) is that it can theoretically use longer wavelengths of light to effect carbon dioxide reduction. The downside to this is that it has a lower driving force for the reaction, and will thus require a better catalyst in order to achieve the same turnover rate. For a recent example of this kind of tradeoff see http://pubs.acs.org/doi/abs/10.1021/ja409267p.

The theoretical energy conversion efficiency will depend on the type of device you are thinking of building. For example, in a dye-sensitized cell the efficiency will be determined by the dye used, with the -0.35 V conduction band enabling the use of dyes which absorb at longer wavelengths.

The redox potential for selective reaction depend the reduction of C-H and C-O bonds which is involve free energy plausible to bond cleavage through single electron or multi electron participation. In the Fermi level of conduction band equilibrium can tend to change the redox potential and that related to the reactant species diagonal contact/interaction on the surface of the catalyst. In dry reforming, CO2 has produces number of intermediate species during its reduction; thereby contacts of reactant species free energy presides resonance of Fermi level equilibrium shifts within the conduction band. As you mentioned two potential favor for CO2 reduction, but certainly the intermediate product conversion is entirely differ one from another. So, I agree with answer is depends on your efficiency.

Although  the conduction band of two catalyst satisfy  thermodynamic potential to drive CO2 reduction, the dynamic process of CO2 reduction is different. Therefore,  I also agree with answer is depends on your efficiency.