LSTM autoencoder for anomaly detection - shifts

I am new to the field of deep learning and am trying to implement an LSTM autoencoder for anomaly detection in time series data. I followed a tutorial on YouTube "Time Series Anomaly Detection with LSTM Autoencoders using Keras & TensorFlow 2 in Python" and used the code provided in their Github repository.

However, when applying the code to my own time series data, I encountered an issue. It seems that the results are shifted by a step. The points marked in the image I attached should be considered anomalies, but the anomaly detection algorithm identifies the preceding points as the anomalies instead.

Results: Image may be NSFW.
Clik here to view.

This is the model I used:

model = keras.Sequential()model.add(keras.layers.LSTM(units=128, input_shape=(X_train.shape[1], X_train.shape[2])))model.add(keras.layers.Dropout(rate=0.2))model.add(keras.layers.RepeatVector(n=X_train.shape[1]))model.add(keras.layers.LSTM(units=128, return_sequences=True))model.add(keras.layers.Dropout(rate=0.2))model.add(keras.layers.TimeDistributed(keras.layers.Dense(units=X_train.shape[2])))model.compile(loss='mae', optimizer='adam')history = model.fit(X_train, y_train, epochs=30, batch_size=5, validation_split=0.1, shuffle=False)

I'm unsure whether the issue stems from the data, the model, or if the observed shifts in anomaly detection are normal and they just need to be shifted back?