multi time step forecasting for next 24 H ahead using XGBoost

Question

My project is multi time step forecasting for next 24 H ahead using XGBoost.

The below two figures are the results:

Why the test data as in first fig, is different than the second fig (orange one), it should be the same?

from sklearn.metrics import mean_squared_error, mean_absolute_error
import numpy as np

# Inverse transform the predictions and the test data (if scaled)
y_pred_actual = scaler_y.inverse_transform(y_pred.reshape(-1, 1)) # Ensure shape is 
(-1, 
   1)
  y_test_inv = scaler_y.inverse_transform(y_test)

# Calculate metrics
mse = mean_squared_error(y_test_inv, y_pred_actual)
mae = mean_absolute_error(y_test_inv, y_pred_actual)
rmse = np.sqrt(mse)
mpe = np.mean((y_test_inv - y_pred_actual) / y_test_inv) * 100

# Calculate MAPE
mape = np.mean(np.abs((y_test_inv - y_pred_actual) / y_test_inv)) * 100

# Print the evaluation metrics
print('')
 print('---------------------------------------------------')
print(f'TL_model_loaded MAE for test set : {round(mae, 3)}')
print(f'TL_model_loaded MSE for test set : {round(mse, 3)}')
print(f'TL_model_loaded RMSE for test set : {round(rmse, 3)}')
print(f'TL_model_loaded MPE for test set : {round(mpe, 3)} %')
print(f'TL_model_loaded MAPE for test set : {round(mape, 3)} %')
print('---------------------------------------------------')
print('')

# Call the plot_results_xgboost function
plot_results_xgboost(y_pred_actual, y_test_inv, evals_result, 'XG')