When, during the process of producing a song, should one "normalise"? Is it wise to normalise everything straight away or at some other time (like at the end of recording a certain instrument)?

The reason I ask is this. Imagine I record one guitar sequence. I chop it in the sequencer and have 2 audio sequences for guitar on one track which originally came from the same recording. If I normalise the first, the whole of the first sequence gets "lifted" in relation to the loudest peak in that file. And ditto the second file. However, the loudest peak in one or the other files might (will probably) be a different level than the other. Thus I end up with different loudnesses.

What do readers of this forum recommend?

thx