Benchmark Scorecard — recall and precision against GIAB HG002, by caller and variant type

"1.0 / 1.0 recall and precision" is the headline number, but the counts behind it are what actually earns trust: this is hap.py's true-positive, false-positive, and false-negative tally for each caller, split by SNP and INDEL, restricted to the padded CAPN3/DMD regions and benchmarked against the GIAB HG002 NISTv4.2.1 truth set.