The filter will trap down to the absolute minimum value, whatever that happens to be. But an ICP will clog the inlet way below that value. ICP measures elements in solution, not particles. Particles are bad for the ICP plasma and are filtered out by the machine.
You still measure Si but it’s (obviously) not what’s being trapped by the oil filter. There’s a gap in sizes between what is trapped and what appears on the analysis, you’ll have particles in the oil that are neither trapped nor appear on the UOA.
FWIW this is for all elements not just silicon.