diff --git a/src/pynxtools/dataconverter/helpers.py b/src/pynxtools/dataconverter/helpers.py
index d9664a967..b3c2e6fe8 100644
--- a/src/pynxtools/dataconverter/helpers.py
+++ b/src/pynxtools/dataconverter/helpers.py
@@ -24,7 +24,7 @@
 from datetime import datetime, timezone
 from enum import Enum
 from functools import lru_cache
-from typing import Any, Callable, List, Optional, Tuple, Union, Sequence, cast
+from typing import Any, Callable, List, Optional, Sequence, Tuple, Union, cast
 
 import h5py
 import lxml.etree as ET
@@ -77,7 +77,14 @@ def __init__(self):
         self.data = set()
         self.logging = True
 
-    def _log(self, path: str, log_type: ValidationProblem, value: Optional[Any], *args):
+    def _log(
+        self,
+        path: str,
+        log_type: ValidationProblem,
+        value: Optional[Any],
+        *args,
+        **kwargs,
+    ):
         if value is None:
             value = "<unknown>"
 
@@ -153,9 +160,13 @@ def _log(self, path: str, log_type: ValidationProblem, value: Optional[Any], *ar
             logger.warning(f"The attribute {path} will not be written.")
         elif log_type == ValidationProblem.InvalidConceptForNonVariadic:
             value = cast(Any, value)
-            log_text = f"Given {value.type} name '{path}' conflicts with the non-variadic name '{value}'"
-            if value.type == "group":
-                log_text += f", which should be of type {value.nx_class}."
+            log_text = (
+                f"Defined non-variadic name '{value.name}' of {value.type} "
+                f"conflicts with the given variadic concept '{path}'."
+            )
+            if "expected_type" in kwargs:
+                log_text = f"{log_text[:-1]} {kwargs['expected_type']}. Such conflict may break pynxtools writer."
+
             logger.warning(log_text)
 
     def collect_and_log(
diff --git a/src/pynxtools/dataconverter/validation.py b/src/pynxtools/dataconverter/validation.py
index d0bd645e2..9e3b7534d 100644
--- a/src/pynxtools/dataconverter/validation.py
+++ b/src/pynxtools/dataconverter/validation.py
@@ -134,7 +134,9 @@ def split_class_and_name_of(name: str) -> Tuple[Optional[str], str]:
     ), f"{name_match.group(2)}{'' if prefix is None else prefix}"
 
 
-def best_namefit_of(name: str, nodes: Iterable[NexusNode]) -> Optional[NexusNode]:
+def best_namefit_of(
+    name: str, nodes: Iterable[NexusNode], expected_type: str
+) -> Optional[NexusNode]:
     """
     Get the best namefit of `name` in `keys`.
 
@@ -166,13 +168,20 @@ def best_namefit_of(name: str, nodes: Iterable[NexusNode]) -> Optional[NexusNode
                         and len(type_attr) > 2
                     ]
                     if concept_name not in inherited_names:
-                        if node.type == "group":
+                        if node.type == "group" and node.type != expected_type:
                             if concept_name != node.nx_class[2:].upper():
                                 collector.collect_and_log(
                                     concept_name,
                                     ValidationProblem.InvalidConceptForNonVariadic,
                                     node,
                                 )
+                        elif node.type == "field" and node.type != expected_type:
+                            collector.collect_and_log(
+                                concept_name,
+                                ValidationProblem.InvalidConceptForNonVariadic,
+                                node,
+                                expected_type=expected_type,
+                            )
                         else:
                             collector.collect_and_log(
                                 concept_name,
@@ -589,12 +598,29 @@ def handle_unknown_type(node: NexusNode, keys: Mapping[str, Any], prev_path: str
         pass
 
     def add_best_matches_for(key: str, node: NexusNode) -> Optional[NexusNode]:
-        for name in key[1:].replace("@", "").split("/"):
+        is_last_attr = False
+        key_components = key[1:].split("/")
+        if "@" in key_components[-1]:
+            is_last_attr = True
+            key_components[-1] = key_components[-1].replace("@", "")
+        key_len = len(key_components)
+        expected_type = "group_or_field"
+        for ind, name in enumerate(key_components):
             children_to_check = [
                 node.search_add_child_for(child)
                 for child in node.get_all_direct_children_names()
             ]
-            node = best_namefit_of(name, children_to_check)
+            if ind < (key_len - 2):
+                expected_type = "group"
+            elif ind == (key_len - 2) and not is_last_attr:
+                expected_type = "group"
+            elif ind == (key_len - 2) and is_last_attr:
+                expected_type = "group_or_field"
+            elif ind == (key_len - 1) and not is_last_attr:
+                expected_type = "field"
+            elif ind == (key_len - 1) and is_last_attr:
+                expected_type = "attribute"
+            node = best_namefit_of(name, children_to_check, expected_type)
 
             if node is None:
                 return None
diff --git a/src/pynxtools/dataconverter/writer.py b/src/pynxtools/dataconverter/writer.py
index f2a289503..cc175609c 100644
--- a/src/pynxtools/dataconverter/writer.py
+++ b/src/pynxtools/dataconverter/writer.py
@@ -249,8 +249,12 @@ def ensure_and_get_parent_node(self, path: str, undocumented_paths) -> h5py.Grou
 
             attrs = self.__nxdl_to_attrs(parent_path)
 
-            if attrs is not None:
+            if attrs is not None and "type" in attrs.keys():
                 grp.attrs["NX_class"] = attrs["type"]
+            else:
+                logger.warning(
+                    f"Nexus group {parent_path_hdf5} is written without NX_class attribute."
+                )
             return grp
         return self.output_nexus[parent_path_hdf5]