Genome-scale metabolic network reconstructions (GENREs) are valuable tools for understanding microbial metabolism. The process of automatically generating GENREs includes identifying metabolic reactions supported by sufficient genomic evidence to generate a draft metabolic network. The draft GENRE is then gapfilled with additional reactions in order to recapitulate specific growth phenotypes as indicated with associated experimental data. Previous methods have implemented absolute mapping thresholds for the reactions automatically included in draft GENREs; however, there is growing evidence that integrating annotation evidence in a continuous form can improve model accuracy. There is a need for flexibility in the structure of GENREs to better account for uncertainty in biological data, unknown regulatory mechanisms, and context-specificity associated with data inputs. To address this issue, we present a novel method that provides a framework for quantifying combined genomic, biochemical, and phenotypic evidence for each biochemical reaction during automated GENRE construction. Our method, Constraint-based Analysis Yielding reaction Usage across metabolic Networks (CANYUNs), generates accurate GENREs with a quantitative metric for the cumulative evidence for each reaction included in the network. The structuring of CANYUNs allows for the simultaneous integration of three data inputs while maintaining all supporting evidence for biochemical reactions that may be active in an organism. CANYUNs is designed to maximize the utility of experimental and annotation datasets and to ultimately assist in the curation of the reference datasets used for the automatic construction of metabolic networks. We validated CANYUNs by generating an E. coli K-12 model and compared it to the manually curated reconstruction iML1515. Finally, we demonstrated the use of CANYUNs to build a model by generating an E. coli Nissle CANYUNs model using novel phenotypic data that we collected. This method may address key challenges for the procedural construction of metabolic networks by leveraging uncertainty and redundancy in biological data.