This paper studies how biomechanical multibody models of scoliosis can neglect the changes of spinal length and yet be accurate in reconstructing spinal columns. As these models with fixed length comprise rigid links interconnected by rotary joints, they resemble polygonal chains that approximate spine curves with a finite number of line segments. In mathematics, using more segments with shorter length can result in more accurate curve approximations. This raises the question of whether more accurate spine curve approximations by increasing the number of links/joints can yield more accurate spinal column reconstructions. For this, the accuracy of spine curve approximation was improved consistently by increasing the number of links/joints, and its effects on the accuracy of spinal column reconstruction were assessed. Positive correlation was found between the accuracy of spine reconstruction and curve approximation. It was shown that while increasing the accuracy of curve approximations, the representation of scoliosis concavity and its side-to-side deviations were improved. Moreover, reconstruction errors of the spine regions separated by the inflection vertebrae had minimal impacts on each other. Overall, multibody scoliosis models with fixed spinal length can benefit from the extra rotational joints that contribute towards the accuracy of spine curve approximation. The outcome of this study leads to concurrent accuracy improvement and simplification of multibody models; joint-link configurations can be independently defined for the regions separated by the inflection vertebrae, enabling local optimization of the models for higher accuracy without unnecessary added complexity to the whole model.