if one has a good experience with callipers, it is not bad. if you want a total safe measurement you have to look for some hight tech stuff like dexa or other super duper methods like water weighing etc. 8 to 10 % is like a good visible six pack with visible veins on your stomach i would say. for further advice ask mark hunt or roy nelson.
DEXA is not that accurate itself. DPA (dual photon absorptiometry) is better. DEXA was more accurate when it was pencil beam geometry but that's dog slow so they went to fan beam geometry sacrificing scatter but still good enough for bone mineral measurements.
But let's take a step back and note two things:
1. The first aim of DEXA is precision, and not accuracy.You want to track change in bone mineral density.Yes - they do compare against a database and give you a Z-score (where you sit on the database distribution), but the real aim is tracking change with age, disease or therapy.
2. the purpose of the fat estimate is to correct for BMC calculation, not the accuracy of fat content in each measurement.
Sure, Hologic, Norland and GELunar report fat because it is one of the calculations that pops out of the methodology but that doesn't mean it's accurate.
AFAIK, water dunk test is the most accurate short of melting down a cadaver.
Dual-energy CT or spectral CT should be damn good but no one is going to take that radiation dose for a body fat measurement.
MRI should be very good since MR is so sensitive to fat that you often have to run a fat suppression sequence.
