Fix reconstruction

I noticed a bug in the reconstruction of the time posterior when using distance marginalisation that was introduced in !535 (merged).

Basically, the wrong thing is being passed to the interpolant so the time recovered is meaningless.