I don't think it says that.
It says that more than three weeks ago only 7.3pc showed an antibody response.
It also says that by *now* they expected 20pc to have it.
Thoughts:
A. That's quite a long interval.
B. Antibody responses take a while to develop.
C. The maths wouldn't have had to be very wrong to account for that difference.
D. Everyone expects the maths to be a bit wrong anyway. We don't have enough data to parameterise the models.
E. These figures are about the same as Barcelona (with caveats of timing differences etc).
I'm not overly concerned although it does show, again, that going for a quick, relatively painless, herd immunity is not a good strategy.