Computed tomography (CT) is a non-invasive technique that allows the detection and classification of the internal structure of an object. However, in several applications, the high doses of radiation generated by the CT scanner significantly increase the risk of damaging the object of interest. To reduce this damage, optimized hardware settings have been proposed by lowering the number of angles at which projections are taken. However, the reduction of measurements leads to a highly ill-pose inverse problem, sensitive to measurements and modeling errors. Coded aperture X-ray tomography is one approach that can overcome these limitations. The cone-beam architecture for computed tomography obtains 3D images of the complete object of interest, thus reducing the time it is exposed to radiation. In this paper we investigate and test sampling strategies for a cone-beam architecture in compressive computed tomography, in this way employing fewer measurements than expected from the classical sampling theory without a significant loss of information. These strategies are tested for a cone-beam architecture, whereas previous approaches were developed in a fan-beam architecture. The results indicate that by using just 25% of the samples, it is possible to obtain from 26 dB until 50 dB in the reconstructed images.