This is something I've been thinking about lately. You look up any reconstructed proto-language, you'll probably find an estimate as to when it was spoken. But how can this number be determined?
From my research, these numbers came from an outdated theory that languages all change at the same rate (which has clearly been proven false), thus you can date a proto-language by the number of changes that have occurred.
Well, if this method has long since been debunked, then why are the results of it still widely accepted? That doesn't make any logical sense. Why would you continue to accept the results decades after the method used to determine said results has been debunked?