The "relationally correct" way to accomplish this is to consider either each begin/end pair or each event as a separate thing. The schema/structure that you've created makes the computation much more difficult than it needs to be.
As an example, if you consider each pair to be a "unit", then you get something like:
If you get more precise and consider each event discreetly, then you need to write a bit more complex SQL but the concept remains the same.
CREATE TABLE Meine_Zeit_Paar (
innie DATETIME NOT NULL
, outie DATETIME NOT NULL
SELECT TIME_FORMAT(Sum(TIMEDIFF(eind, start)),'%H:%i') AS gewerkt
By using a simpler model, you get simpler and more robust SQL that copes with unanticipated issues better.
In theory, theory and practice are identical. In practice, theory and practice are unrelated.