It had no measurable effect in the measured group. You can't say that it works based on that. If it had a measurable effect in that group, you would expect to see improvement in the treatment group between the two scenarios.
If (65-69 no treatment) == (70-74 no treatment) and also (65-69 no treatment) == (70-74 treatment), then (70-74 treatment) == (70-74 no treatment) by the transitive property of equality.
Now the reason for (65-69 no treatment) == (70-74 no treatment) may be its own mystery, but we expect to see at least _something_ happen anyway between (70-74 no treatment) and (70-74 treatment) if the treatment had some benefit for that group, and apparently they didn't see that.
There was no significant different before or after Paxlovid became available to the early 70s patients. This updates us against Paxlovid being effective in this patient population, for the specific outcome metrics.