(1) falsifying research data
(2) concealing, deleting or otherwise destroying emails, information or data
(3)misusing privileged information and
(4)seriously deviating from accepted practices for proposing, conducting or reporting research and other scholarly activities.
The NSF undertook its own investigation because, as a federal body, it has its own standards and wasn't initially given the data by the university, which found no evidence for the first three concerns and investigated and cleared Mann for item 4 and concluded no wrongdoing.
The NSF instead found items 2-4 were okay with their own findings corroborating the university and had concerns about item 1 instead, but their investigation said no falsification had occurred. Their conclusion did note
There are several concerns raised about the quality of the statistical analysis techniques that were used in the Subject's researchand
there was concern about how extensively the Subject's research had influenced the debate in the overall research field.but mistakes are a part of science and it isn't a fault his work is well regarded and so he is popular.
Much of the current debate focuses on the viability of the statistical procedures he employed, the statistics used to confirm the accuracy of the results, and the degree to which one specific set of data impacts the statistical results. These concerns are all appropriate for scientific debate and to assist the research community in directing future research efforts to improve understanding in this field of research. Such scientific debate is ongoing but does not, in itself, constitute evidence of research misconduct.They're right, it isn't misconduct. Given a do-over, the 'statistical analysis techniques' they didn't like - the hockey stick cobbling together temperature data and then tree rings to visually demonstrate a continued upward curve in temperature - did a lot more harm than good so I am sure he would not do it again either but science moves on, and now so can Mann.