I mean, the FPTP system is fucked – that aside... For your stat to be right, wouldn't you have to calculate it also in terms of the number of constituents to a given constituency? E.g., Constituency A has 10 constituents, Constituency B has 15 constituents, and Constituency C has 100 constituents; both A + B have a non-Lab/Cons party in first or second place; C has Lab first and Cons second. In that scenario, it wouldn't be true that "most people" live in a constituency where Lab or Cons are not both 1st or 2nd (where A-C is exhaustive).
I don't know how that would extrapolate to the real constituencies with their varying population/electorate figures. Certainly, it's a very uneven and strange system at present, which allows for all sorts of gerrymandering. But only given the sum of 20 constituencies (per your calculation) where Lab/Cons are not the top two, I don't think you can infer the situation for most people.
In any case: fuck this system, PR soon please.