Quality in the capital allocation process is of utmost relevance in securing and sustaining economic performance and corporate goals. The correct characterisation of uncertainties across the opportunity set is critical in achieving optimal capital allocation decisions. Real projects are characterised by distributional forms that are of a very different nature to those found in securities markets, demanding the utilisation of a more general measure of risk. This paper demonstrates the significant differences in portfolio optimisation outcomes as a result of using variance and semivariance as alternative measures of risk. It is shown that the use of semivariance leads to a more robust efficient frontier. The results have broad implications with regard to the choice of a suitable risk measure within the general context of real projects portfolio optimisation. The proper selection of a risk measure can lead to significant improvements in the quality of decisions with regard to capital allocation.