Forecast with newer data do not uses existing values

### Operating System

Windows

### Version Information

Python Version 3.8

### Steps to reproduce

In sdk/python/jobs/pipelines/1h_automl_in_pipeline/automl-forecasting-in-pipeline/automl-forecasting-in-pipeline.ipynb at Chapter 4.1 it is explained, that one can update the data with newer values and leave NaNs for the values that should be predicted. However when I try to do it, it just overwrites my actual values and not just the NaNs... Can someone explain how that is possible?

### Expected behavior

The actual values should stay how they are, only NaNs should be overwritten.

### Actual behavior

Actual values are overwritten with predicted values.

### Addition information

_No response_

Start with sdk/python/jobs/pipelines/1h_automl_in_pipeline/automl-forecasting-in-pipeline/automl-forecasting-in-pipeline.ipynb, especially Chapter 4.1, and run the described update with newer values and NaNs. Trace how the forecasting output is applied to the updated data; done means existing actual values remain unchanged while only NaNs receive predictions.

Written by the indexing model from the issue text.