‘TIOBE’ does appear to be based on "verifiable (or at least falsifiable) figures"...
No one's been able to reproduce them. (They confused a web browser with a language implementation for several months!) They're based on bad assumptions. No one's done TIOBE better because it's meaningless to do the wrong things better. See TIOBE or not TIOBE – “Lies, damned lies, and statistics” for one explanation.
Also, unlike science, they don't release raw data sets.
Even if you believe in tiobe, this graph does not support JavaFan's taunt that "the haydays of Perl where in the mid to late 90s" [sic]. That graph indicates that "Perl's tiobe haydays" were 2003-2005. At least that's what I see. So I doubt JavaFan had tiobe in mind when making his "haydays" wisecrack.