It seems the paper made mistakes, using a non-standard model of the CMB that failed upon replication. It's an interesting theory thoughb.