One question to ask yourself to figure out which system you like better is what the outcome should be of this election:
Candidate 1 is loved by 55% of the population and despised by 45%. Candidate 2 is loved by 45% of the population and despised by 55%. Candidate 3 is tolerated by 85%.
Approval Voting would choose Candidate 3. IRV would choose Candidate 1. Who "should" be elected?
http://scorevoting.net/UtilFoundns.html
So the right answer is presumably #3, although you didn't specify exact utilities so I don't know for sure.
Everyone gets much better representation when "their candidate" can actually get a seat at the table, instead of having one party/candidate "take it all" (in that local region of theirs). It also spurs collaboration [2] and "bridge-building" rather than division and obstruction as it happens with the FPTP in the U.S. (and elsewhere)
My biggest hope is that Canada is closing-in on a proportional representation voting system, and adopts it within the next year. That could (or should) make it an election issue for 2018 and 2020 in the U.S., too.
Numerical scale rankings have similar problems (especially as there is empirical evidence that numerical rankings for similar actual levels of satisfaction are assigned differently in different cultures (e.g., ethnic) even within, say, the US. Again, there are circumstances other than usual candidate voting where the numerical scale can be tied to a concrete commitment to make the meaning of the markings consistent and such voting useful (e.g., when they are tied to an amount that one would pay to get or avoid a particular outcome), but they are generally a bad idea for electing candidates. Ranked ballots is pretty much the system that provides the most information with relatively consistent meaning from each voter for circumstances like public, secret-ballot, free elections of candidates. (Different election methods using ranked ballots do a better or worse job of using that information, of course.)
> Approval voting is a single-winner voting method used for elections. Each voter may "approve" of (i.e., select) any number of candidates. The winner is the most-approved candidate.
It's worth noting that there is a mathematical discipline of studying voting systems. Approval has been studied quite extensively and seems to have fewer flaws than most other methods—though it has been proven that no method can be perfect.