Oryza sativa (rice) plays an essential food security role for more than half of the world's population. Obtaining crops with high levels of disease resistance is a major challenge for breeders, especially today, given the urgent need for agriculture to be more sustainable. Plant resistance genes are mainly encoded by three large leucine-rich repeat (LRR)-containing receptor (LRR-CR) families: the LRR-receptor-like kinase (LRR-RLK), LRR-receptor-like protein (LRR-RLP) and nucleotide-binding LRR receptor (NLR). Using lrrprofiler, a pipeline that we developed to annotate and classify these proteins, we compared three publicly available annotations of the rice Nipponbare reference genome. The extended discrepancies that we observed for LRR-CR gene models led us to perform an in-depth manual curation of their annotations while paying special attention to nonsense mutations. We then transferred this manually curated annotation to Kitaake, a cultivar that is closely related to Nipponbare, using an optimized strategy. Here, we discuss the breakthrough achieved by manual curation when comparing genomes and, in addition to 'functional' and 'structural' annotations, we propose that the community adopts this approach, which we call 'comprehensive' annotation. The resulting data are crucial for further studies on the natural variability and evolution of LRR-CR genes in order to promote their use in breeding future resilient varieties.