ECDSA time in Endorsement vs Validation phase

[manavdeep]

Quoting from the paper: https://ieeexplore.ieee.org/document/9912226 :
"It is well-known that validation phase (i.e., validator peer, and not the consensus mechanism which creates blocks) is one of the major bottlenecks and critically affects the achievable peak throughput"

So we have been benchmarking and using pprof library to see the time spent in various phases.
We have 5 organizations, 1 EP and 1 CP each. My policy is and of all Orgs.
We were hoping that the Ecdsa time would be substantially high (around 5x) for the validation phase as compared to endorsing phase.
But that doesn't seem to be the observations. And contradict the general understanding I quoted above from the paper and in general.

What are we doing wrong. Any other way to measure phase wise time instead of pprof?