The attribution of hadron masses to particular contributions is a subtle and not at all intuitive matter. See this question and its answers for extended discussion on where hadron mass "comes from". Saying
"the valence quarks make up 1,5%" and attributing the rest to "sea particles"
is simply wrong - the "rest" is a complex combination of different energies, which are related to "virtual particles" only insofar as that - like most QFT computations - virtual particles appear when you try to compute these perturbations perturbatively. In particular about 1/4th of the mass comes from the "trace anomaly", whose nature you cannot even explain in any naive picture of just a bunch of particles bound together.
The "sea" of particles inside a hadron refers to the parton model that gives us parton distribution functions for which particles we are likely to observe when we probe a hadron at a certain energy scale. Crucially, these are a function of the energy of the probe, i.e. what we expect to find "inside" a hadron depends on how we're looking at it. Since quantum chromodynamics is strongly interacting at every day energies, perturbation theory (and hence by extension the concept of virtual particles) is useless in understanding how hadrons that are just sitting around work.