The study appears to have made no effort to compare counties that reopened dining to those (in California?) which did not. Rather it simply took each county and compared numbers before and after restrictions were lifted. Considering that most of the lifting of restrictions occurred late Summer and into Fall, that means that the study effectively attributes the Winter surge to the lifting of dining restrictions.
Similarly, the decline from mask mandates doesn't compare mandate counties to non-mandate counties (despite there now being ample non-mandate states to compare). It simply compares pre-mandate numbers in a county to post-mandate numbers. Considering the mandates were mostly imposed during the initial wave(s), it effectively credits the mandates for declines from peaks without any additional analysis.
Given the varying levels of restrictions across different US (and International) jurisdictions, it seems like a more compelling analysis could have been made. But it appears that there is little correlation between level of NPIs and case/death numbers. If there was a clear causal relationship, then Florida and some other Southern states would have seen clearly worse results than more restrictive states. But that hasn't been the case.