Well, I imagine a lot of those casualties were not from traps but from engaging the enemy.
But let's pretend the 33% is solely from traps. Which means 66% were never badly injured or killed by traps. Since those traps are designed to seriously wound or kill, and soldiers don't have "hit points" that heal in 24 hours, I am guessing that it only takes one bad mistake....maybe two if you're lucky....to get taken out of the war.
Assuming that "only" 33% of tunnel rats got wounded or killed by traps, it's pretty easy to do some math showing the relationship between the number of traps encountered, and the frequency at which they were avoided. Or not avoided.
For example, if on average a tunnel rat encounters only 10 traps total, over their course of their deployment, and the chance of detection a trap is N%, then N^10 = 66%, and N is ~96%. That is, if tunnel rats have an 96% chance of detecting a trap, and they encounter a total of 10 traps each, expected survival rate is 66%.
From your description, it sounds like they encounter a lot more traps than that. Let's say 50. And now let's just guess that only half of those casualties are traps, which we'll round up to 17%. So now we need detection rate that, when raised to the 50th power, equals 83%.
Drumroll....
~99.6%
So if we are going for realism, and using Vietnam as our archetype, delvers with expertise in trap detection who are actively looking for traps should have a success rate nearing 100%.