This paper compares the maximum allowable powers of some typical cellular telephones at 835 and 1900 MHz for compliance with the limits of specific absorption rates (SAR) given in ANSI/IEEE, ICNIRP and the proposed modification of ANSI/IEEE safety guidelines. It is shown that the present ANSI/IEEE guideline is the most conservative with the ICNIRP guidelines allowing a maximum radiated powerthat is 2.5-3 times higher, and the proposed IEEE modification of treating pinna as an extremity tissue the least conservative allowing even higher radiated powers by up to 50%. The paper also expands the previously reported study of energy deposition in models of adults versus children to two different and distinct anatomically-based models of the adult head, namely the Utah model and the 'Visible Man' model, each of which is increased or reduced by the voxel size to obtain additional head models larger or smaller in all dimensions by 11.1% or -9.1%, respectively. The peak 1 g body-tissue SAR calculated using the widely accepted FDTD method for smaller models is up to 56% higher at 1900 MHz and up to 20% higher at 835 MHz compared to the larger models, with the average models giving intermediate SARs. Also given in the paper is a comparison of the peak 1 g and 10 g SARs for two different anatomically-based models with 6 mm thick smooth plastic ear models used for SAR compliance testing. The SARs obtained with the insulating plastic ear models are up to two or more times smaller than realistic anatomic models. We propose a 2 mm thin shell phantom with lossy ear that should give SARs within +/- 15% of those of anatomic models.