diff --git a/data/test_data/md5sum.txt b/data/test_data/md5sum.txt
index d227693..0987ae9 100644
--- a/data/test_data/md5sum.txt
+++ b/data/test_data/md5sum.txt
@@ -1,6 +1,6 @@
 e4f4641559aadc9ec7b5c50cc4aa82e3  bam_kmers_out/bam_1/fastqc_data.txt
 090a74739499d287a7aeea689a78b4e9  bam_mult_out/bam_1/fastqc_data.txt
-7cc30a3a14796140f05940485ab1bed5  bam_mult_out/bam_2/fastqc_data.txt
+e3568c45f8cc358c263e51c4ece215cd  bam_mult_out/bam_2/fastqc_data.txt
 090a74739499d287a7aeea689a78b4e9  bam_orig_dups_out/bam_1/fastqc_data.txt
 090a74739499d287a7aeea689a78b4e9  bam_out/bam_1/fastqc_data.txt
 090a74739499d287a7aeea689a78b4e9  bam_tiles_out/bam_1/fastqc_data.txt
diff --git a/src/adapter_matcher.cpp b/src/adapter_matcher.cpp
index bc9552c..e088e15 100644
--- a/src/adapter_matcher.cpp
+++ b/src/adapter_matcher.cpp
@@ -55,13 +55,16 @@ adapter_matcher::add_and_consume(adapter_matcher &&rhs) -> void {
 [[nodiscard]] auto
 adapter_matcher::get_grade(const std::uint64_t n_reads) const -> std::string {
   static constexpr auto label = "adapter";
-  const auto max_count =
-    adap_counts.empty()
-      ? 0
-      : std::ranges::max(adap_counts | std::views::transform(std::ranges::max));
+  // The report shows, per adapter, the cumulative percentage of reads with the
+  // adapter starting at or before each position; as in FastQC, the grade is
+  // the maximum of that, i.e. the total count of each adapter over positions.
+  std::vector<std::uint64_t> totals(n_adapters, 0);
+  for (const auto &counts : adap_counts)
+    std::ranges::transform(totals, counts, std::begin(totals), std::plus{});
+  const auto max_count = totals.empty() ? 0 : std::ranges::max(totals);
   return grader_set::get_grade(
     label,
-    as_frac(max_count, std::max(static_cast<std::uint64_t>(1), n_reads)));
+    pct(as_frac(max_count, std::max(static_cast<std::uint64_t>(1), n_reads))));
 }
 
 [[nodiscard]] auto
diff --git a/src/duplication_results.cpp b/src/duplication_results.cpp
index 0a33242..bfdb405 100644
--- a/src/duplication_results.cpp
+++ b/src/duplication_results.cpp
@@ -157,11 +157,12 @@ duplication_results::add_and_consume(duplication_results &rhs) -> void {
 get_grade_overrepresented(const std::uint64_t n_reads,
                           const duplication_results &dr) -> std::string {
   static constexpr auto label = "overrepresented";
-  const auto max_n_obs =
-    dr.dups.empty() ? 0LU : std::ranges::max(std::views::values(dr.dups));
-  return grader_set::get_grade(
-    label,
-    as_frac(max_n_obs, std::max(static_cast<std::uint64_t>(1), n_reads)));
+  // As in FastQC, grade on the list the report shows: the percentage of the
+  // most frequent listed sequence (in percent, like the cutoffs), or zero if
+  // no sequence passes the list's cutoffs.
+  const auto overrep = dr.get_overrepresented(n_reads);
+  return grader_set::get_grade(label,
+                               overrep.empty() ? 0.0 : overrep.front().pct_val);
 }
 
 [[nodiscard]] auto
@@ -283,7 +284,7 @@ get_grade_duplication(const dup_summary_t &summary) -> std::string {
   };
   const auto frac_dedup =
     as_frac(reduce(summary.hist_dedup), reduce(summary.hist_mass));
-  return grader_set::get_grade(label, frac_dedup);
+  return grader_set::get_grade(label, pct(frac_dedup));
 }
 
 [[nodiscard]] auto
diff --git a/src/falco_grade.cpp b/src/falco_grade.cpp
index 28896e0..7c558e7 100644
--- a/src/falco_grade.cpp
+++ b/src/falco_grade.cpp
@@ -24,15 +24,15 @@ static const auto default_graders = std::array{  // NOLINT(cert-err58-cpp)
   grader{.name = "quality_base_median", .warn = 25.0, .fail = 20.0},
   grader{.name = "quality_base_lower",  .warn = 10.0, .fail = 5.00},
   grader{.name = "quality_sequence",    .warn = 27.0, .fail = 20.0},
-  grader{.name = "n_content",           .warn = 0.05, .fail = 0.20},
+  grader{.name = "n_content",           .warn = 5.00, .fail = 20.0},
   grader{.name = "sequence",            .warn = 10.0, .fail = 20.0},
   grader{.name = "sequence_length",     .warn = 1.00, .fail = 1.00},
   grader{.name = "gc_sequence",         .warn = 15.0, .fail = 30.0},
-  grader{.name = "duplication",         .warn = 0.70, .fail = 0.50},
+  grader{.name = "duplication",         .warn = 70.0, .fail = 50.0},
   grader{.name = "overrepresented",     .warn = 0.10, .fail = 1.00},
   grader{.name = "kmer",                .warn = 2.00, .fail = 5.00},
   grader{.name = "tile",                .warn = 5.00, .fail = 10.0},
-  grader{.name = "adapter",             .warn = 0.05, .fail = 0.10},
+  grader{.name = "adapter",             .warn = 5.00, .fail = 10.0},
 };
 // clang-format on
 
@@ -74,9 +74,11 @@ file_grades::get_title(const std::string &name) -> std::string {
 grader::identify_grade(const double value) const -> std::string {
   using std::string_literals::operator""s;
   if (warn < fail) {
-    if (value < warn)
+    // higher is worse: as in FastQC, a value equal to a cutoff does not
+    // reach it (e.g., 0.1% overrepresented passes, more than 0.1% warns)
+    if (value <= warn)
       return "pass"s;
-    if (value < fail)
+    if (value <= fail)
       return "warn"s;
     return "fail"s;
   }
@@ -198,7 +200,7 @@ get_grade_n_content(const std::vector<std::uint64_t> &n_counts,
     std::reduce(std::cbegin(nucs[max_idx]), std::cend(nucs[max_idx]));
   const auto max_n_frac =
     as_frac(n_counts[max_idx], n_counts[max_idx] + total_non_n);
-  return grader_set::get_grade(label, max_n_frac);
+  return grader_set::get_grade(label, pct(max_n_frac));
 }
 
 [[nodiscard]] auto