diff --git a/src/regtech_data_validator/data_formatters.py b/src/regtech_data_validator/data_formatters.py
index cdb8b7e..62bd790 100644
--- a/src/regtech_data_validator/data_formatters.py
+++ b/src/regtech_data_validator/data_formatters.py
@@ -234,12 +234,12 @@ def df_to_table(df: pl.DataFrame) -> str:
     return tabulate(df, headers='keys', showindex=True, tablefmt='rounded_outline')  # type: ignore
 
 
-def df_to_json(df: pl.DataFrame, max_records: int = 10000, max_group_size: int = 100) -> str:
+def df_to_json(df: pl.DataFrame, max_records: int = 10000, max_group_size: int = 200) -> str:
     results = df_to_dicts(df, max_records, max_group_size)
     return ujson.dumps(results, indent=4, escape_forward_slashes=False)
 
 
-def df_to_dicts(df: pl.DataFrame, max_records: int = 10000, max_group_size: int = 100) -> list[dict]:
+def df_to_dicts(df: pl.DataFrame, max_records: int = 10000, max_group_size: int = 200) -> list[dict]:
     json_results = []
     if not df.is_empty():
         # polars str columns sort by entry, not lexigraphical sorting like we'd expect, so cast the column to use
@@ -264,7 +264,7 @@ def df_to_dicts(df: pl.DataFrame, max_records: int = 10000, max_group_size: int
 # So this function uses the group error counts to truncate on record numbers
 def truncate_validation_group_records(group, group_size):
     need_to_truncate = group.select(pl.col('row').n_unique()).item() > group_size
-    unique_record_nos = group.select('row').unique().limit(group_size)
+    unique_record_nos = group.select('row').unique(maintain_order=True).limit(group_size)
     truncated_group = group.filter(pl.col('row').is_in(unique_record_nos['row']))
     return truncated_group, need_to_truncate
 
diff --git a/src/regtech_data_validator/validator.py b/src/regtech_data_validator/validator.py
index 508d60c..6663fbc 100644
--- a/src/regtech_data_validator/validator.py
+++ b/src/regtech_data_validator/validator.py
@@ -44,7 +44,6 @@ def _get_check_fields(check: Check, primary_column: str) -> list[str]:
 # Retrieves the row data from the original dataframe that threw errors/warnings, and pulls out the fields/values
 # from the original row data that caused the error/warning
 def _filter_valid_records(df: pl.DataFrame, check_output: pl.Series, fields: list[str]) -> pl.DataFrame:
-
     sorted_check_output = check_output["index"]
     fields = ["index"] + fields
     filtered_df = df.filter(pl.col('index').is_in(sorted_check_output))
@@ -69,7 +68,7 @@ def _add_validation_metadata(failed_check_fields_df: pl.DataFrame, check: SBLChe
     return validation_fields_df
 
 
-def validate(schema: pa.DataFrameSchema, submission_df: pl.LazyFrame, process_errors: bool) -> pl.DataFrame:
+def validate(schema: pa.DataFrameSchema, submission_df: pl.LazyFrame, row_start: int, process_errors: bool) -> pl.DataFrame:
     """
     validate received dataframe with schema and return list of
     schema errors
@@ -85,7 +84,7 @@ def validate(schema: pa.DataFrameSchema, submission_df: pl.LazyFrame, process_er
     try:
         # since polars dataframes don't normally have an index column, add it, so that we can match
         # up original submission rows with rows found with errors/warnings
-        submission_df = submission_df.with_row_index()
+        submission_df = submission_df.with_row_index(offset=row_start)
         schema(submission_df, lazy=True)
     except SchemaErrors as err:
         check_findings = []
@@ -187,7 +186,7 @@ def validate_batch_csv(
         # than reading in the whole csv and just selecting on the UID column (currently our only register level check data)
         uids = pl.scan_csv(real_path, infer_schema_length=0, missing_utf8_is_empty_string=True).select("uid").collect()
         register_schema = get_register_schema(context)
-        validation_results = validate(register_schema, uids, True)
+        validation_results = validate(register_schema, uids, 0, True)
         if not validation_results.findings.is_empty():
             validation_results.findings = format_findings(
                 validation_results.findings,
@@ -215,9 +214,10 @@ def validate_chunks(schema, path, batch_size, batch_count, max_errors, checks):
     batches = reader.next_batches(batch_count)
     process_errors = True
     total_count = 0
+    row_start = 0
     while batches:
         df = pl.concat(batches)
-        validation_results = validate(schema, df, process_errors)
+        validation_results = validate(schema, df, row_start, process_errors)
         if not validation_results.findings.is_empty():
             validation_results.findings = format_findings(
                 validation_results.findings, validation_results.phase.value, checks
@@ -230,6 +230,7 @@ def validate_chunks(schema, path, batch_size, batch_count, max_errors, checks):
             head_count = validation_results.findings.height - (total_count - max_errors)
             validation_results.findings = validation_results.findings.head(head_count)
 
+        row_start += df.height
         batches = reader.next_batches(batch_count)
         yield validation_results