Working on the digger definition of a cave, which is anything we could travel through regardless of whether it has yet been discovered or currently has an entrance, I say 1% is in the ballpark for order of magnitude, or even much too high. Of course, that would geologically be considered a conduit, and if you say all caves that currently have entrances, I'd be willing to go higher, to maybe 10%, assuming we are talking about percent of currently accessible passage with no digging required that has been mapped to any degree of accuracy between Grade 3 and Grade 5. On the other hand, I'd have to go back down to 1% if you do percent caves by entrances rather than percent of total cave passage length in the world.
I recall a presentation at the 2010 NSS convention in VT talking about estimating how many caves have been discovered, using a particular county in, I believe, Virginia and starting with an estimate from some 50 years back that the 17 caves in the country represented <10% of the caves total, then contrasting it with many hundreds of caves currently, and rate of cave discovery implying that the new number was actually 1% or less, suggesting the first estimate was too high by orders of magnitude. Maybe whoever did that can comment, because my memory from 4 years ago has likely ret-conned the numbers. Hopefully I'm in the order of magnitude ballpark.