I can’t recall the dtruss equivalent, but were I looking at this my first inclination would be to get an strace with full timing info including the length of time spent in each system call and then start comparing. You don’t say if there’s differing OS versions installed, but changes in the runtime, standard libraries, third-party but linked in libs, yadda yadda yadda . . . . Handwaving for sure but ballpark two hundredths of a second difference doesn’t strike me as completely outside the realm of what I’d consider a reasonable “expected variation”.

Edit: and I’m basing that .02 difference as .74 vs .72 rounded