diff --git a/R/frs_cluster.R b/R/frs_cluster.R
index 5f4cdc0..3f1d349 100644
--- a/R/frs_cluster.R
+++ b/R/frs_cluster.R
@@ -382,11 +382,21 @@ frs_cluster <- function(conn, table, habitat,
 }
 
 
-#' Both-direction cluster connectivity check
+#' Both-direction cluster connectivity check (three-phase)
 #'
-#' Evaluates upstream and downstream independently on the original data,
-#' then combines: a cluster is valid if connected in either direction.
-#' Uses a temp table for clustering so both checks see the same clusters.
+#' Phase 1: segments with BOTH `label_cluster` and `label_connect` TRUE
+#' are always valid (on-spawning rearing). These are excluded from
+#' clustering so they don't dilute cluster boundaries.
+#'
+#' Phase 2: upstream boolean — rearing clusters below spawning. Uses
+#' `FWA_Upstream()` to check if spawning exists anywhere upstream.
+#'
+#' Phase 3: downstream trace — rearing clusters above spawning. Uses
+#' `FWA_Downstream()` on the broken streams table (mainstem only) with
+#' gradient bridge and distance cap.
+#'
+#' A cluster is valid if it passes Phase 2 OR Phase 3. Phase 1 segments
+#' are always valid regardless of clustering.
 #'
 #' @noRd
 .frs_cluster_both <- function(conn, table, habitat,
@@ -401,7 +411,19 @@ frs_cluster <- function(conn, table, habitat,
   tmp_clusters <- sprintf("pg_temp.frs_clusters_%s",
     gsub("[^a-z0-9]", "", tolower(species)))
 
-  # Create shared clustering temp table
+  # Phase 1: on-spawning segments — always valid, excluded from clustering
+  phase1_ids <- DBI::dbGetQuery(conn, sprintf(
+    "SELECT h.id_segment FROM %s h
+     WHERE h.species_code = %s AND h.%s IS TRUE AND h.%s IS TRUE",
+    habitat, sp_quoted, label_cluster, label_connect))$id_segment
+
+  # Cluster remaining (non-phase1) segments
+  phase1_filter <- if (length(phase1_ids) > 0) {
+    sprintf("AND h.id_segment NOT IN (%s)", paste(phase1_ids, collapse = ", "))
+  } else {
+    ""
+  }
+
   .frs_db_execute(conn, sprintf("DROP TABLE IF EXISTS %s", tmp_clusters))
   .frs_db_execute(conn, sprintf(
     "CREATE TEMP TABLE %s AS
@@ -415,14 +437,14 @@ frs_cluster <- function(conn, table, habitat,
      FROM %s h
      INNER JOIN %s s ON h.id_segment = s.id_segment
      WHERE h.species_code = %s
-       AND h.%s IS TRUE",
-    tmp_clusters, habitat, table, sp_quoted, label_cluster))
+       AND h.%s IS TRUE %s",
+    tmp_clusters, habitat, table, sp_quoted, label_cluster, phase1_filter))
 
   .frs_db_execute(conn, sprintf(
     "CREATE INDEX ON %s (cluster_id)", tmp_clusters))
 
-  # Upstream valid clusters — boolean check (spawning exists anywhere upstream)
-  sql_upstream <- sprintf(
+  # Phase 2: upstream boolean — rearing below spawning
+  sql_phase2 <- sprintf(
     "WITH cluster_minimums AS (
        SELECT DISTINCT ON (cluster_id)
          cluster_id, wscode_ltree, localcode_ltree,
@@ -454,10 +476,12 @@ frs_cluster <- function(conn, table, habitat,
     habitat, sp_quoted, label_connect, table,
     conf_m)
 
-  valid_up <- DBI::dbGetQuery(conn, sql_upstream)$cluster_id
+  valid_phase2 <- DBI::dbGetQuery(conn, sql_phase2)$cluster_id
 
-  # Downstream valid clusters
-  sql_downstream <- sprintf(
+  # Phase 3: downstream trace on broken streams table — rearing above spawning
+  # Uses FWA_Downstream on the working streams table (not fwa_downstreamtrace)
+  # Mainstem only (blue_line_key = watershed_key) for linear path
+  sql_phase3 <- sprintf(
     "WITH cluster_minimums AS (
        SELECT DISTINCT ON (cluster_id)
          cluster_id, wscode_ltree, localcode_ltree,
@@ -469,33 +493,29 @@ frs_cluster <- function(conn, table, habitat,
      downstream AS (
        SELECT
          cm.cluster_id,
-         t.linear_feature_id,
-         t.wscode,
-         t.downstream_route_measure,
-         t.gradient,
-         EXISTS (
-           SELECT 1 FROM %s s
-           INNER JOIN %s h2 ON s.id_segment = h2.id_segment
-           WHERE s.linear_feature_id = t.linear_feature_id
-             AND h2.species_code = %s
-             AND h2.%s IS TRUE
-         ) AS has_connect,
-         -t.length_metre + SUM(t.length_metre) OVER (
+         st.gradient,
+         st.wscode_ltree,
+         st.downstream_route_measure,
+         COALESCE(h2.%s, false) AS has_connect,
+         -st.length_metre + SUM(st.length_metre) OVER (
            PARTITION BY cm.cluster_id
-           ORDER BY t.wscode DESC, t.downstream_route_measure DESC
+           ORDER BY st.wscode_ltree DESC, st.downstream_route_measure DESC
          ) AS dist_to_cluster
        FROM cluster_minimums cm
-       CROSS JOIN LATERAL whse_basemapping.fwa_downstreamtrace(
-         cm.blue_line_key,
-         cm.downstream_route_measure
-       ) t
-       WHERE t.blue_line_key = t.watershed_key
+       INNER JOIN %s st ON FWA_Downstream(
+         cm.blue_line_key, cm.downstream_route_measure,
+         cm.wscode_ltree, cm.localcode_ltree,
+         st.blue_line_key, st.downstream_route_measure,
+         st.wscode_ltree, st.localcode_ltree)
+       LEFT JOIN %s h2 ON st.id_segment = h2.id_segment
+         AND h2.species_code = %s
+       WHERE st.blue_line_key = st.watershed_key
      ),
      downstream_capped AS (
        SELECT
          row_number() OVER (
            PARTITION BY cluster_id
-           ORDER BY wscode DESC, downstream_route_measure DESC
+           ORDER BY wscode_ltree DESC, downstream_route_measure DESC
          ) AS rn,
          *
        FROM downstream
@@ -505,29 +525,30 @@ frs_cluster <- function(conn, table, habitat,
        SELECT DISTINCT ON (cluster_id) *
        FROM downstream_capped
        WHERE has_connect IS TRUE
-       ORDER BY cluster_id, wscode DESC, downstream_route_measure DESC
+       ORDER BY cluster_id, wscode_ltree DESC, downstream_route_measure DESC
      ),
      nearest_barrier AS (
        SELECT DISTINCT ON (cluster_id) *
        FROM downstream_capped
        WHERE gradient >= %s
-       ORDER BY cluster_id, wscode DESC, downstream_route_measure DESC
+       ORDER BY cluster_id, wscode_ltree DESC, downstream_route_measure DESC
      )
      SELECT a.cluster_id
      FROM nearest_connect a
      LEFT JOIN nearest_barrier b ON a.cluster_id = b.cluster_id
      WHERE b.rn IS NULL OR b.rn > a.rn",
     tmp_clusters,
-    table, habitat, sp_quoted, label_connect,
+    label_connect,
+    table, habitat, sp_quoted,
     bd,
     bg)
 
-  valid_down <- DBI::dbGetQuery(conn, sql_downstream)$cluster_id
+  valid_phase3 <- DBI::dbGetQuery(conn, sql_phase3)$cluster_id
 
-  # Union of both valid sets
-  all_valid <- unique(c(valid_up, valid_down))
+  # Union: valid in Phase 2 OR Phase 3
+  all_valid <- unique(c(valid_phase2, valid_phase3))
 
-  # UPDATE: set FALSE for segments in clusters NOT valid in either direction
+  # UPDATE: set FALSE for clustered segments NOT valid in either phase
   if (length(all_valid) == 0) {
     .frs_db_execute(conn, sprintf(
       "UPDATE %s h SET %s = FALSE