Most of beamforming methods rely on the proper estimation of the steering vectors of the incoming signals in order to calculate appropriate array feeding weights. However, this estimation is often performed on a theoretical basis, since most methods assume that the array elements radiate like isotropic sources and consider only the phase differences between them due to their different spatial positions. This is a rough estimation as it ignores both the non-isotropic radiation pattern of the array elements and the mutual coupling between them. To overcome these limitations, we use a different approach for the construction of steering vectors based on the embedded patterns of the array elements, extracted through full-wave analysis. This approach is used in this paper to modify two well-known beamforming methods, namely the null steering beamforming and the minimum variance distortionless response method. Both modified methods are applied to a realistic model of a microstrip planar antenna array, and thus, both the azimuth and polar angles of the main lobe and nulls directions are controlled (beamforming in 3D space). Extensive statistical analyses are performed by implementing several scenarios of escalating difficulty, in order to examine the performance and temporal response of the modified beamforming methods in comparison to the respective conventional methods and also in comparison to beamforming methods based on evolutionary techniques.