Using Prophet for Time Series gap imputation

Consider a regular hourly time series with a relatively long sequence of consecutive missing values or a gap.

I'm considering using Prophet for the following approach:
- Identify start and end of gap
- Apply linear interpolation
- Create a left and right window 2x the size of the gap
- Fit this series in Prophet (left window, center window or linear interpolated gap, right window)
- Predict the same period with Prophet

**Example**
Sample Dataset: ETTh1.csv
Corrupted series: 1000 hour gap
The plot of the result (original, corrupted, reconstructed)
MAE of linear interpolation: 2.79
MAE of Prophet: 2.44

**In theory does this approach make sense?**
Notebook: https://colab.research.google.com/drive/1w7y5ju0hnd0zLp6qTwmW7ukLgLozvW-b?usp=sharing