Researcher - Formating Issues

Solved. After many experiments, I figured out that the prompt and the instructions must be well visually identified and captured (e.g. instructions separated with line breaks and bullets or grouped). When I asked copilot for this, it verified this empirical finding. The causality is that large language models can partially fail when prompted with compacted texts paying most attention to the begining or end of the prompt only. Very impressive and not self explained. The response to the formating instructions was outstanding when I re-engineered the prompt so that to provide spot instructions visually separated with line breaks and segments. Hope this will help others (it took me 1 week intensive trial and error testing)