Exactly as much difference as it should make 
. With only 5 winners, “proportionality” doesn’t really come into play unless at least 20% of the voters are under-represented in some way. This can be so in a “highly visible” way if voters can be partitioned into relatively large sets voting for nearly mutually exclusive sets of candidates. Else it’s subtler.
In our test case, the top 3 candidates had 15, 14, and 11 approvals, all super-majorities. Any sane form of approval voting was bound to select all 3. They were in a class of their own.
Next came one candidate with 9 approvals, and 5 candidates tied at 8 approvals each.
So every candidate save one was approved by at least half the ballots. There’s no “gross injustice” in sight.
Still, I found it impressive that PAV cut through the 5-way tie to settle on a unique winning set, maximizing as best it could the “satisfaction” (by one family of measures) across all ballots.
I like that much better than, e.g., flipping a crypto-coin to break a tie - but not enough better to justify the considerable pain of trying to switch to a different voting service. And PAV can end up with tied winning sets too (although it didn’t in this specific case).