Fix overconfident statement

tools/templates/README.tmpl

@@ -153,11 +153,14 @@ This is a (temporary) fork of
 
 ### Data Quality & Fidelity
 
-This dataset achieves **100% fidelity** for linguistic data (tokens, annotations, dependencies) and **~99.98% fidelity** for metadata. The Parquet files can be perfectly reconstructed back to the original CoNLL-U format with:
-- ✅ All linguistic annotations preserved exactly
-- ✅ Multi-word tokens (MWTs) and empty nodes fully supported
-- ✅ Duplicate metadata keys preserved (1,323 sentences across 14 treebanks)
-- ✅ Enhanced dependencies and rare annotation edge cases handled correctly
+This dataset achieves **100% fidelity** for linguistic data (tokens,
+annotations, dependencies) and **very high (~98%) fidelity** for metadata. The
+Parquet files can be perfectly reconstructed back to the original CoNLL-U
+format with:
+- All linguistic annotations preserved exactly
+- Multi-word tokens (MWTs) and empty nodes fully supported
+- Duplicate metadata keys preserved (1,323 sentences across 14 treebanks)
+- Enhanced dependencies and rare annotation edge cases handled correctly
 
 Recent improvements include fixes for:
 - Double equals parsing in FEATS/MISC fields (e.g., `Gloss==POSS`)