The negative adjustments of past temperatures are neither the work of one scientist, nor even one scientific organization. Let's say for simplicity that they're the result of ten entirely separate confirmed hypotheses about systematic errors in old data.
Each of these was produced by an honest researcher who followed the scientific method: formulate a hypothesis, test it, etc. Let's say their technical work is perfect. All the hypotheses are correct.
But all ten researchers, faced with the choice of what hypothesis to formulate and/or investigate, naturally chose high-impact hypotheses -- ones that, if confirmed, adjust past temperatures downward.
We would naturally expect historical errors in temperature measurements to be evenly distributed. Like errors in cheap thermometers. If they are evenly distributed, we would expect there to be ten other such hypotheses, involving upward corrections in past measurements, which were not studied because of their low impact if confirmed.
This hypothesis explains the data that we ourselves see: that past temperatures are continuously being lowered as the data evolves. And it requires no one to be a bad, incompetent, or dishonest researcher.