diff --git a/CHANGELOG.md b/CHANGELOG.md
index d80836ca..086c68ec 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -12,7 +12,8 @@
 ### 🗑️ Removed
 - Remove Python 3.9 support ([#457](https://github.com/cbrnr/mnelab/pull/457) by [Clemens Brunner](https://github.com/cbrnr))
 
-### 🔧 Fixed
+### Fixed
+- Fix an issue where some .xdf files were not loaded correctly (wrong dtype or multi-channel marker) ([#464](https://github.com/cbrnr/mnelab/pull/464) by [Benedikt Klöckl](https://github.com/bkloeckl))
 - Fix a bug where appending data would not be correctly displayed in the history ([#446](https://github.com/cbrnr/mnelab/pull/446) by [Benedikt Klöckl](https://github.com/bkloeckl))
 - Fix resetting the settings to default values ([#456](https://github.com/cbrnr/mnelab/pull/456) by [Clemens Brunner](https://github.com/cbrnr))
 - Fix an issue where the channel montage figure could not be closed on macOS ([#459](https://github.com/cbrnr/mnelab/pull/459) by [Benedikt Klöckl](https://github.com/bkloeckl)) 
diff --git a/src/mnelab/dialogs/xdf_streams.py b/src/mnelab/dialogs/xdf_streams.py
index 21208370..aacb81c6 100644
--- a/src/mnelab/dialogs/xdf_streams.py
+++ b/src/mnelab/dialogs/xdf_streams.py
@@ -151,7 +151,7 @@ def selected_streams(self):
         for row in self.view.selectionModel().selectedRows():
             type_ = self.view.item(row.row(), 2).text()
             fs = self.view.item(row.row(), 5).value()
-            if type_ != "Markers" and fs != 0:
+            if not _is_marker(type_) and fs != 0:
                 streams.append(self.view.item(row.row(), 0).value())
         return streams
 
@@ -160,7 +160,11 @@ def selected_markers(self):
         markers = []
         for row in self.view.selectionModel().selectedRows():
             type_ = self.view.item(row.row(), 2).text()
-            fs = self.view.item(row.row(), 5).value()
-            if type_ == "Markers" and fs == 0:
+            if _is_marker(type_):
                 markers.append(self.view.item(row.row(), 0).value())
         return markers
+
+
+def _is_marker(type):
+    marker_types = ["Markers", "Marker", "StringMarker"]
+    return type in marker_types
diff --git a/src/mnelab/io/xdf.py b/src/mnelab/io/xdf.py
index a954d57d..12594461 100644
--- a/src/mnelab/io/xdf.py
+++ b/src/mnelab/io/xdf.py
@@ -34,6 +34,7 @@ def __init__(
             Resampling target frequency in Hz. If only one stream_id is given, this can
             be `None`, in which case no resampling is performed.
         """
+
         if len(stream_ids) > 1 and fs_new is None:
             raise ValueError(
                 "Argument `fs_new` is required when reading multiple streams."
@@ -42,7 +43,7 @@ def __init__(
         streams, _ = load_xdf(fname)
         streams = {stream["info"]["stream_id"]: stream for stream in streams}
 
-        if all(_is_markerstream(streams[stream_id]) for stream_id in stream_ids):
+        if (not stream_ids) and marker_ids:
             raise RuntimeError(
                 "Loading only marker streams is not supported, at least one stream must"
                 " be a regular stream."
@@ -53,6 +54,23 @@ def __init__(
         for stream_id in stream_ids:
             stream = streams[stream_id]
 
+            # check if the dtype is valid, try convertion otherwise
+            dtype = stream["time_series"].dtype
+            if dtype not in [np.float64, np.complex128]:
+                try:
+                    stream["time_series"] = stream["time_series"].astype(np.float64)
+                except ValueError:
+                    try:
+                        stream["time_series"] = stream["time_series"].astype(
+                            np.complex128
+                        )
+                    except ValueError as e:
+                        raise RuntimeError(
+                            f"Stream {stream['info']['name']} has unsupported"
+                            " dtype {dtype}. "
+                            f"Conversion to float64 and complex128 failed."
+                        ) from e
+
             n_chans = int(stream["info"]["channel_count"][0])
             labels, types, units = [], [], []
             try:
@@ -92,17 +110,23 @@ def __init__(
         super().__init__(preload=data, info=info, filenames=[fname])
 
         # convert marker streams to annotations
-        for stream_id, stream in streams.items():
-            if marker_ids is not None and stream_id not in marker_ids:
-                continue
-            if not _is_markerstream(stream):
-                continue
+        for marker_id in marker_ids:
+            stream = streams[marker_id]
+            channel_count = int(stream["info"]["channel_count"][0])
             onsets = stream["time_stamps"] - first_time
-            prefix = f"{stream_id}-" if prefix_markers else ""
-            descriptions = [
-                f"{prefix}{item}" for sub in stream["time_series"] for item in sub
-            ]
-            self.annotations.append(onsets, [0] * len(onsets), descriptions)
+            prefix = f"{marker_id}-" if prefix_markers else ""
+
+            if channel_count == 1:
+                # handle single-channel markers
+                descriptions = [f"{prefix}{item[0]}" for item in stream["time_series"]]
+                self.annotations.append(onsets, [0] * len(onsets), descriptions)
+            else:
+                # handle multi-channel markers
+                for sample in stream["time_series"]:
+                    for _, description in enumerate(sample):
+                        self.annotations.append(
+                            onsets, [0] * len(onsets), str(description)
+                        )
 
 
 def _resample_streams(streams, stream_ids, fs_new):
@@ -196,12 +220,6 @@ def read_raw_xdf(
     return RawXDF(fname, stream_ids, marker_ids, prefix_markers, fs_new)
 
 
-def _is_markerstream(stream):
-    srate = float(stream["info"]["nominal_srate"][0])
-    n_chans = int(stream["info"]["channel_count"][0])
-    return srate == 0 and n_chans == 1
-
-
 def get_xml(fname):
     """Get XML stream headers and footers from all streams.