DNA-guided cell-free protein synthesis using a minimal set of purified components has emerged as a versatile platform in constructive biology. The E. coli-based PURE (protein synthesis using recombinant elements) system offers the basic protein synthesis factory in a prospective minimal cell relying on extant molecules. However, there is an urgent need to improve the system's performance and to build a mechanistic computational model that can help interpret and predict gene expression dynamics. Herein, we utilized all three commercially available PURE system variants: PURExpress, PUREfrex and PUREfrex2.0. We monitored apparent kinetics of mRNA and protein synthesis by fluorescence spectroscopy at different concentrations of DNA template. Analysis of polysome distributions by atomic force microscopy, combined with a stochastic model of translation, revealed inefficient usage of ribosomes, consistent with the idea that translation initiation is a limiting step. This preliminary dataset was used to formulate hypotheses regarding possible mechanisms impeding robust gene expression. Next, we challenged these hypotheses by devising targeted experiments aimed to alleviate the current limitations of PUREfrex. We identified depletion of key initiation factors (IFs) by translationally inactive mRNA as a possible inhibitory mechanism. This adverse process could partly be remedied by targeted mRNA degradation, whereas addition of more IFs and of the hrpA RNA helicase had no substantial effects. Moreover, the depletion of tRNAs as peptidyl-tRNAs can become limiting in PUREfrex (but not in PURExpress), which can be alleviated by addition of peptidyl-tRNA-hydrolase (PTH). We attempted to build a new model for PURE system dynamics integrating all experimental observations. Although a satisfying global fit can be obtained in specific conditions (with PTH), a unifying system's level model is still missing.