A succession dispute arose following the death of Edward the Confessor, King of England in 1066. William, Duke of Normandy ultimately prevailed following the demise of his main competitor at the Battle of Hastings. Was the Anglo-Saxon state a distinct entity from the Anglo-Norman one? Edward the Confessor never gets a numeral, so I'm wondering why Edward Longshanks is Edward I and not Edward II of England.
Some regnal numbers are applied retrospectively, but as you suggest they begin at around the time of William the Conqueror and their introduction from Frankish tradition, although they're not consistently applied.
The reason for some retrospective application is that accepted histories of royal lineage also changed - for many centuries the work of Geoffrey of Monmouth was considered an authoritative view of the line despite the existence of numerous ficitious monarchs in his work. By the 18th century much of Geoffrey's work had been disproved and revised, and it's in that period that we see some simultaneous revision of royal ordinals.
Another reason for the emergence of ordinal useage is a narrowing in the number of families providing monarchs and a demise in the use of sobriquets. For example, one Henry Tudor is easily confused with another while Erik Bloodaxe and Erik the Red are easily told apart in histories. You also have to consider useage on a case-by-case basis, Edward the Confessor is remembered by sobruiquet and was therefore already identifiable without an ordinal. One might feel that the "rule" has been misapplied there but in fact there wasn't really a rule, it was simply a way of telling who was who in the histories.
There were no "Firsts" in their own lifetimes, that was always applied retrospectively once a "Second" was on the throne, and often it was even later than that, so Edward I (named, we're told, after Edward the Confessor) was just plain ol' Edward until there was an Edward II.
Sources:
Geoffrey of Monmouth, Jankulak K, 2000,
The Frankish World, Nelson J, 1966
Uniting the Kingdom, Grant A, Stringer K, 2002