Really? I must be thinking of something else then (or it was an oversimplification). The theory given in my high school chemistry was
1) All atoms want 8 valence electrons, except for the smallest.
2) Non metals share electrons so they increase the number around them to 8
3) Metals give up electrons so their inner 8 surround them.
Given that, P is 3 electrons short, we'd predict it would form PF3. Noble gases have 8, so there's no reason for them to form a bond.
You have confused the octet rule with VSEPR (and in fact the octet rule does not work for all the first row elements, cf boron). The octet rule would predict that phosphorus forms a PF3 complex, and indeed it does. However, phosphorus is a second row element and can expand its octet to give PF5. There are 5 bonding pairs around the central atom, VSEPR predicts a trigonal bipyramidal geometry, precisely in accord with the solution and solid state structures of PF5. On the other hand VSEPR predicts a pyramidal structure for PF3 again in accord with actuality.
Valence Shell Electron Pair Repulsion theory is a very simple idea; it takes bonding and non-bonding electron pairs about a central atom, and arranges them into Platonic solids to predict geometry. Despite it simplicity, there are very few coumpounds of the main group whose structures cannot be predicted by VSEPR. Thanks to pgwenthold for the heads up about XeF3, that would be a potent oxidant, I think I'll pass on handling that.