Bloggers aren't accurate!

Or more correctly, there exists some bloggers who are inaccurate.
This is almost a rerun of my posts Computers aren't accurate, which I made last week. It turns out that the study I used as an example of computers 'believing' the wrong outcome is probably cooked, and maybe as far back as 1952.
This was pointed out to me by Milan Ninchich (as well as being suggested by Mark Weeks of chess.about.com). Firstly, White needs to hang onto the f pawn (contrary to my initial suggestion) as otherwise the King will be driven into the corner in a similar manner to the 'box' method in KQvK mates. If that happens the white bishop on f7 is forced to move, and white will lose wither the bishop of the g pawn, and eventually the game. So the soundness of the study revolves around whether White can hang onto the f pawn.
Some discussion/analysis on the rybka board shows that White cannot do so, meaning that Black wins. After 1.g5+ Kh7 2.Bf7 c2 3.Kh5 c1(Q) 4.h6+ Kh8 5.Kg4 Black plays 5. ... Qc3! and will eventually win the f pawn. The analysis (not mine mind you) can be found at the Rybka link.