Probably has to do with their goal seek based data analysis and raw data noise reduction, where noise is defined as cooler measured temperatures (you know the absence of the heat values they need for the goal seek). They just massage inn some extra heat in those areas and kneed the averages until the temperature rises appropriately for the result they are seeking.

Temperatures average planet wide is at best the same and more likely on average falling slowly. Things peeked in the late 90s on legitimately measuring temperature rise (and the measured rise back then was negligible then anyway).

Carbon also is not a significant factor in planetary temperature given it's small percentage of the whole atmosphere. The sun however is the major factor and the large ocean bodies and their multi ocean current belts also play a significant role.