I think I've seen the graph you are referring to from Imperial College - it is reproduced in this article.
https://www.spectator.co.uk/article/...efore-lockdown
How did they know? It must have been pure guesswork since as WP points out there were so few tests being carried out. What we can now do is look at hospital admissions and deaths and work backwards. This is what Bristol University has done to calculate that peak infection was before lockdown and that the R rate was coming down quickly by the day of lockdown.
Logically the Imperial College Graph makes no sense since it implies that all the social distancing actions people were taking in the days before 24th March had next to no effect on the rate of infection. How likely is that?
As the article says, the Norwegians also believe that their infections had peaked before lockdown.