The reason people think the Chess.com rating system is more accurate, is because it had used a more standardized default rating, and now the average on the site is closer to that of FIDE, USCF, etc. In reality, ratings are arbitrary numbers, and the only way ratings can be compared is with other ratings in the same system (site, federation). Chess.com actually uses the archaic Glicko-1, as opposed to the newer Glicko-2, which has important modifications in the realm of volatility, meaning the chance a player does something unexpected, and it leads to more accurate ratings, not less. Lichess, on the other hand, has updated to Glicko-2.