diff --git a/crates/sds-core/src/pathfind.rs b/crates/sds-core/src/pathfind.rs
index 076ba5f..d350d82 100644
--- a/crates/sds-core/src/pathfind.rs
+++ b/crates/sds-core/src/pathfind.rs
@@ -647,15 +647,17 @@ mod tests {
         assert_eq!(three.target_movement_modifier(), 1);
     }
 
-    /// Pinned counts on a board with nothing on it, and on one that is all
-    /// heavy woods. A change in either is a change in the search, and has to be
-    /// explained rather than absorbed.
+    /// Pinned counts on two **uniform synthetic boards**: one 48x48 with
+    /// nothing on it, one 48x48 that is all level-2 woods. A change in either
+    /// is a change in the search, and has to be explained rather than absorbed.
     ///
-    /// The `candidates` epic quotes 41 hexes and 151 states for 6 MP on open
-    /// ground. This search says 66 and 210 from the same budget; the epic's
-    /// figures were measured before this code existed, and
-    /// `bridge/sds/SdsPathfind.java` - MegaMek's own answer - is what settles
-    /// the difference.
+    /// **Not comparable to the figures in the `candidates` epic.** Those were
+    /// measured on the real suite sheets, `Map Set 5/16x17 Open Terrain 1` and
+    /// `Map Set 4/16x17 Heavy Forest 1`, which carry mixed terrain, elevation
+    /// and edges a 6 MP mech walks into. An idealised plain reaches further
+    /// than a real open-terrain sheet, so a larger number here contradicts
+    /// nothing. These boards are uniform on purpose: they make the count a
+    /// function of the search and of nothing else.
     #[test]
     fn counts_are_pinned() {
         let open_board = open(48, 48);
diff --git a/plan/candidates.md b/plan/candidates.md
index e3d916a..07f0f78 100644
--- a/plan/candidates.md
+++ b/plan/candidates.md
@@ -30,8 +30,17 @@ hexside of turning:
 
 | board | reachable hexes | reachable `(hex, facing)` states |
 |---|---|---|
-| open terrain | 41 | 151 |
-| heavy forest | 10 | 37 |
+| `Map Set 5/16x17 Open Terrain 1` | 41 | 151 |
+| `Map Set 4/16x17 Heavy Forest 1` | 10 | 37 |
+
+Both are suite boards, with mixed terrain, elevation and edges a 6 MP mech
+reaches. A uniform synthetic board is a different measurement and its counts do
+not belong in this table: 6 MP on an all-clear plain reaches more than a real
+open-terrain sheet does, because the sheet has terrain and edges in it.
+
+These figures give 3.68 and 3.70 facings per hex, which contradicts the
+2.6-3.5 band recorded further down. Both are pre-corpus figures and both stand
+as written until the differential corpus settles which is right.
 
 So the scoring work - line of sight, hit locations, damage distributions,
 per-location value - is being used to rank five options out of scores. **No
@@ -94,11 +103,11 @@ good offence and bad defence, and a single operator over `N` cannot say so.
       offline fixture, committed once**, the way the LOS corpus and the hit
       tables are - nothing dumps during a match. Bound it to a handful of
       `(board, start, MP)` combinations
-- [x] Whatever movement inputs the board message lacks - terrain kind,
-      elevation - added to the wire. Cover is *not* among them: the LOS map
-      already carries it. The board already carried both; what it could not
-      carry is the *unit*, so `movementMode` and `maxElevationChange` went on
-      the observation instead
+- [x] Whatever movement inputs the wire lacks, added to it. The board message
+      already carried terrain kind and elevation, and the LOS map already
+      carries cover; what nothing carried is the *unit*, so `movementMode` -
+      which column of `Terrain.movementCost` applies - and
+      `maxElevationChange` went on the observation
 
 **Ground movement**