Keeping images breaks parsing

[eugeniobaglieri]

I’m testing trafilatura 2.0.0 (CLI and Python) and noticed unexpected behavior with this URL:
https://ricette.giallozafferano.it/Gnocchi-con-crema-di-zucchine.html

Running:

 trafilatura -u https://ricette.giallozafferano.it/Gnocchi-con-crema-di-zucchine.html --markdown

produces a consistent extraction: without_images.md
However, adding --images changes the output significantly:

trafilatura -u https://ricette.giallozafferano.it/Gnocchi-con-crema-di-zucchine.html --markdown --images

Results in:

• different and incomplete content extraction
• markdown formatting is not respected anymore
• images are still not included

with_images.md