Non-coding gene regulatory loci are essential to transcription in mammalian cells. As a result, a large variety of experimental and computational strategies have been developed to identify cis-regulatory enhancer sequences. However, in practice, most studies consider enhancer candidates identified by a single method alone. Here we assess the robustness of conclusions based on such a paradigm by comparing enhancer sets identified by different strategies. Because the field currently lacks a comprehensive gold standard, our goal was not to identify the best identification strategy, but rather to quantify the consistency of enhancer sets identified by ten representative identification strategies and to assess the robustness of conclusions based on one approach alone. We found significant dissimilarity between enhancer sets in terms of genomic characteristics, evolutionary conservation, and association with functional loci. This substantial disagreement between enhancer sets within the same biological context is sufficient to influence downstream biological interpretations, and to lead to disparate scientific conclusions about enhancer biology and disease mechanisms. Specifically, we find that different enhancer sets in the same context vary significantly in their overlap with GWAS SNPs and eQTL, and that the majority of GWAS SNPs and eQTL overlap enhancers identified by only a single identification strategy. Furthermore, we find limited evidence that enhancer candidates identified by multiple strategies are more likely to have regulatory function than enhancer candidates identified by a single method. The difficulty of consistently identifying and categorizing enhancers presents a major challenge to mapping the genetic architecture of complex disease, and to interpreting variants found in patient genomes. To facilitate evaluation of the effects of different annotation approaches on studies' conclusions, we developed a database of enhancer annotations in common biological contexts, creDB, which is designed to integrate into bioinformatics workflows. Our results highlight the inherent complexity of enhancer biology and argue that current approaches have yet to adequately account for enhancer diversity.