This articles explain well why it’s not really possible for any full time working adult to learn the tone of each word passively.

A word (in any language really) uttered mid sentences deviates from citation form.

Chinese speakers do not ignore tones when speaking longer sentences, but they bend them within a range that is acceptable to their native comprehension.

On the graph, I see a word here or there overlap in pitch and contour that belong to different tones at different placements in an utterance, nevertheless, it’s only acceptable where the prosodic environment permit, something that is not up to 2L learners to decide whether it’s permissible, let alone a single individual.