Purpose: A post‐processing dual‐energy CT (pDECT) method for non‐invasive estimation of photon cross‐section information has been experimentally implemented on a commercial CT scanner and an initial accuracy assessment performed in an idealized phantom geometry. Methods: Eight test materials of known composition were used to compare pDECT‐estimated linear attenuation coefficients to NIST reference values over the energy range of 10‐1,000keV. CT sinogram data was acquired on a commercial fan beam CT scanner at 90 and 140kVp. Improvement from increased spectral separation was also investigated by additional filtration of the high‐energy beam with 0.5mm of tin. pDECT accuracy was compared using images reconstructed with a conventional filtered backprojection (FBP) algorithm and an in‐house statistical iterative reconstruction algorithm, Alternating Minimization (AM), since AM has been shown to reconstruct images with less random and systematic error than FBP. Results: Mean accuracy of pDECT‐estimated linear attenuation coefficients, assessed by averaging pixels (N>149) within the test material, were found to be within 1% of reference for energies between 30‐1,000 keV, with errors rising to 3––6% at 10 keV. However, the sensitivity of the pDECT process to reconstructed image uncertainties leads to large pixel‐to‐pixel variations of estimated attenuation coefficients; in some cases a coefficient of variation on the order of 20% is seen. The AM algorithm noise advantage leads to attenuation coefficient standard deviations roughly 40% to 60% less than FBP. Measurements using increased spectral separation returned similar mean accuracy and slightly less pixel‐to‐pixel variation even when scanning with 25% of the unfiltered dose. Conclusion: The advantage of the AM algorithm over FBP in this idealized pDECT scenario is less random uncertainty. Increased spectral separation better conditions the pDECT problem. Future work intends to investigate pDECT accuracy in more complex phantom geometries where AM reconstruction may confer more advantage due to its more physically accurate forward model. This work was supported in part from grants (R01 CA 75371 and R01 CA149305, J. Williamson, Principal Investigator) awarded by the National Institutes of Health and a grant funded by Varian Medical Systems.