If you're sure the game runs with the chips in those locations then you have a labelling error. Those are the original stickers too so rather strange to have this happen.
The ROMs must be installed in their correct locations. If the socket for IC17 has the wrong chip installed, the game will not boot. If one or more of the other ROM sockets doesnt have the correct chip, the game may boot, but will definitely not function correctly if it does boot.
If you suspect it's NOT a labelling error you can swap those two ROMs, put them in their correct locations, and see if your board boots in the game. This will not hurt anything. I would say this is quite likely the issue with that board.
As for your original question, the tests are designed to run in order so that you can verify certain aspects of the game (like displays etc) because the later tests use more functionality (like displays etc). However, if you think you need to jump to another test down the line, then it doesnt hurt anything to do so and then jump back. It just depends on what your theory of the problem is and whether you think jumping to a later test for a moment will help you prove or disprove that theory.
You could run all of the tests on the bench or in the game - you just have to be aware of the limitations of either scenario. Running a test that shows results on the displays won't be very useful on the bench unless you have a set of displays connected on the bench.
I would swap those ROMs and install the boards into the game and see if it boots into attract mode. Either way you could then finish the remainder of the test ROMs in game just for the sake of completion. You may find other smaller issues that you wouldn't otherwise find out until after playing the game for a while.