Understand Pelt detected change points

[AC-Meira]

Hello.

I need a help to undestand why the Pelt algorithm sometimes detects a point in the middle of two segments and sometimes detects two points, one when the change from one segment to the other began, and another point when the change ended. The image below shows what I tried to describe:

You can clearly see that the series has 3 different segments, but the Pelt algorithm detected 4 change points instead of 2.

I noticed that when there are fewer points in the serie, the Pelt algorithm detects fewer points. For example, in the image below, the serie has fewer points (~120 points, and the example above has ~200 points) and the algorithm has detected only 1 change point between the first and second segments:

Additional information:
Pelt parameters: model='rbf', cost='rpt.costs.CostRbf()', min_size=1, jump=1.
To predict and find the best pen parameter, I'm using this function:
point_detection_penalty = 15
x_lines = algo.predict(pen=point_detection_penalty)
while point_detection_penalty >= len(x_lines):
point_detection_penalty -= 1
x_lines = algo.predict(pen=point_detection_penalty)
if len(x_lines) > 15:
x_lines = x_lines[-1:]
return x_lines

I would like the algorithm to detect only one point between the different segments of the series. Can someone help with that, please?

Thanks in advance.