The x in your first equation is the number of people you check; the x in your second equation(s) is the number (or rather, proportion of the total number) of the people you succeed in determining the gender of. So, the system isn't valid; this isn't that kind of problem.
To identify the sex of all the people that are identifiable - i.e. to solve the problem as stated, you will have to run all of the tests on every person. This will take 1:40 if the tests can't be run in parallel, and 1:00 if they can.
However, you can optimize the process in various ways depending on your goals. If you want to do something with the people you determine the sex of, but only have a limited rate capacity for determined-person processing, you should run the 10-second test first, so you can have people to process ASAP and not waste post-processing capacity.
If, on the other hand, you want to make it so more people are found proportionally in the early part of the process than the late part, you can see that the hairstyle gives 1%/s, shoes give .67%/s, and jewelry gives 2%/s. So again, you should start with jewelry (and continue with hairstyle), although the fact that the different processes fall into the same order based both on time taken per test and time-efficiency is not applicable to a general case.