Accurate detection of antimicrobial resistance genes (ARGs) from sequencing data is vital for surveillance, clinical diagnostics, and research. Databases such as CARD and ResFinder enable rapid in silico sequence-to-phenotype profile prediction of microbiome sequence data. However, these analyses are often undermined by limitations in [meta]genomic assembly and gene prediction. Consequently, many genuine ARGs, especially low-abundance variants, remain undetected, as nucleotide level variation is frequently lost. Existing ARG detection tools exacerbate this by clustering database sequences using arbitrary thresholds, imposing hidden parameter defaults, and obscuring intermediate results behind summarised outputs. We present AMRfíor (pronounced AMR Feer), a transparent, multi-tool workflow that integrates BLAST, DIAMOND, Bowtie2, BWA and Minimap2 to interrogate both DNA and protein sequences against multiple AMR databases, including any user-provided database. By enforcing user-definable and gene-centric coverage validation, AMRfíor mitigates false positives while capturing true (fíor - Irish for ‘true’), often-missed variants. AMRfíor represents a back-to-basics approach at ARG detection, where direct read-gene mapping/alignment was able to identify SNP-level ARG variation in multiple [meta]genomic samples. Additionally, by comparing the outcomes of DNA and protein sequence search techniques, we are able to observe environmental- and species-specific variants of ARGs, which could provide insight into resistome adaptation.  By exposing detection-tool parameters, allowing customisable ARG detection thresholds and providing both complete mapping statistics and tool outputs, we can observe how both user and developer choices, such as search algorithms, databases, and parameters, can dramatically alter the reported resistomes, influencing study conclusions.