The analogy works just fine if you really do see what I'm getting at.
You can't use your car's odometer to measure your drapes because it's only calibrated to tenths of a mile. Since it can only measure to the nearest tenth of a mile, anything less than a tenth of a mile will return... a tenth of a mile. Every set of drapes you measure will be a tenth of a mile, so there's no point in measuring at all. Just go to the store and order "drapes".
Similarly, the K-Ar method is calibrated to - let's say - millions of years. Since it can only measure to the nearest million years, anything less than a million years will return... a million years. Every young sample you measure will be a million years, so there's no point in measuring at all.
That doesn't mean that either your odometer or K-Ar dating is inaccurate.
Don't take the exact numbers too seriously - I don't know what the actual sensitivity of K-Ar is, off hand. But the principle is sound and the analogy is sound.
If that's not the objection you have, then please be more clear in expressing it.