I was hoping Melanie would come to my defense about the reverse polarity diode.
You don't need me to defend you Adam *smiles* just asbestos underpants...

Your original circuit was good - as long as it's a fuseable Resistor (only problem they're expensive). Most amateurs wouldn't know what one of them was so they'd be fitting normal Resistors... then you've got a potential flame-out in a vehicle... mounted next to the flexible plastic piece of gasoline hose they've just fitted whilst parked at the filling station...

Geee... I think I missed my calling - should be writing disaster movie scripts in Hollywood...

For cheap and effective protection I willingly sacrifice a bit of accuracy (or compensate in software). The resistor divider values can be reduced and a better choice of zenner to minimise the effect.
It's not a 'bit' of accuracy, it's a shed-load of it. There's no way of calculating other than manually calibrating your product. Zeners aren't consistent. 5.6v Zener is probably +/- 5% (or even 10%) - well you do the math. To protect the PICs ADC input you need a 5.6v Zener max. That could start conducting at about 3v... therefore your useable ADC input on a PIC is between 0 and 3v tops! Above 3v it's going to be inaccurate because of that Zener. Better to have a larger Resistor Ratio, plan for things going wrong... plan for some twerp putting your device across a 24v truck battery and calculate your Resistor chain accordingly.

For those that are wondering what I'm talking about, and have been taught at college that Zeners conduct at their rated voltage, put your DVM set to mA in series with one and wind a PSU up slowly (via a current limiting Resistor). You will discover they start conducting long before you actually expect them to. Believe nothing - trust no-one (especially college lecturers!)... whatever happened to Mulder & Scully?...