docs: correct the reproduction instructions

DATA.md

@@ -37,8 +37,10 @@ idempotent. Verify before training:
 python modeling/rebase_graph_paths.py --csv <csv> --check     # expect N/N
 ```
 
-Expected: BACH 14,341, BreakHis 709, BRACS 96,153, SPIDER-breast 71,745,
-TCGA-BRCA 11,149,499. Anything less means graphs are being silently skipped.
+`--check` stops after 20,000 rows and says so (`(first 20k rows only)`), so it prints the
+full count only for BACH (14,341) and BreakHis (709). For BRACS, SPIDER-breast and
+TCGA-BRCA it prints `20000/20000`; that is the pass condition for those three. A
+resolvable count below the total means graphs are being silently skipped.
 
 ## Graph format
 
@@ -93,20 +95,42 @@ python modeling/rebase_graph_paths.py --csv graphs/bach/extracted/sample_labels_
 BRACS, 7-class lesion subtyping. 6 × ~1 GB tars under `bracs/data/` holding **only** `.pt`
 files. Unlike BACH, the `sample_labels_rich.csv` sits one level up at `bracs/`.
 
+```bash
+mkdir -p graphs/bracs/extracted
+for t in graphs/bracs/data/*.tar; do tar -xf "$t" -C graphs/bracs/extracted; done
+python modeling/rebase_graph_paths.py --csv graphs/bracs/sample_labels_rich.csv \
+       --mode absolute --root "$PWD/graphs/bracs/extracted"
+```
+
 ### `graphs/breakhis/`: 709 graphs
 BreakHis, binary benign/malignant. Small enough to ship as loose `.pt` files under
-`breakhis/graphs/`, with `sample_labels_rich.csv` alongside them.
+`breakhis/graphs/`, with `sample_labels_rich.csv` alongside them. Nothing to extract, but
+the CSV still needs rebasing:
+
+```bash
+python modeling/rebase_graph_paths.py --csv graphs/breakhis/graphs/sample_labels_rich.csv \
+       --mode absolute --root "$PWD/graphs/breakhis/graphs"
+```
 
 ### `graphs/spider_breast/`: 71,745 graphs
-SPIDER-breast, 18 tumour-subtype classes, train 62,995 / test 8,750. 18 × ~245 MB tars under
-`spider_breast/data/` (`.pt` only; the CSV is one level up).
+SPIDER-breast, 18 tumour-subtype classes, train 62,995 / test 8,750. 18 × ~255 MB tars under
+`spider_breast/data/` (`.pt` only; the CSVs are one level up).
+
+```bash
+mkdir -p graphs/spider_breast/extracted
+for t in graphs/spider_breast/data/*.tar; do tar -xf "$t" -C graphs/spider_breast/extracted; done
+python modeling/rebase_graph_paths.py --csv graphs/spider_breast/sample_labels_rich.csv \
+       --mode absolute --root "$PWD/graphs/spider_breast/extracted"
+```
 
 Two things to know:
 
 - **The label files cover 92,892 patches, the graphs are 71,745.** Graph construction skips
   any patch with fewer than 10 detected cells, which removes cell-sparse tissue almost
   entirely: `Fat` retains 7 of 6,286 patches (1 in test), `Fibrosis` 415 of 6,260,
-  `Necrosis` 2,147 of 5,396, while ten other classes lose ~0 %. In an 18-class macro-F1 this
+  `Necrosis` 2,147 of 5,396, `Lipogranuloma` 2,622 of 4,941 and `Vessels` 3,130 of 5,469.
+  Two more lose a tenth or less (`Benign phyllodes tumor` 89.9 %, `Fibroadenoma` 93.7 %);
+  the remaining eleven classes retain at least 97 %. In an 18-class macro-F1 this
   matters, because `Fat` rests on a single test patch. `metadata.csv` is the file that matches the
   graphs exactly.
 - **`sample_id` is doubled in `metadata.csv`** (`patch_0000017_patch_0000017`, matching the
@@ -358,7 +382,9 @@ python studies/survival/investigate.py \
   --input_csv survival.csv --output_dir out/survival --model_type graphist
 ```
 
-Swap `--embeddings_dir` for any directory under `baselines/` to get the comparison rows.
+Swap `--embeddings_dir` for `baselines/dinov2`, `baselines/mae` or
+`baselines/graphist_v1/embeddings` to get the comparison rows. Note the extra level on the
+last one: `baselines/graphist_v1/` itself holds the v1 checkpoint, not the embeddings.
 Requires `lifelines`.
 
 **SPIDER-breast patch-level probe.** Same as slide-level but with `--scale patch` and