GPS accuracy

This question is motivated by the recent discussion about deriving
polars from flight logs.

We know that GPS fixes have some amount of (absolute) error. What is
their accuracy relative to each other? In other words, when I'm
analyzing flight logs, how accurately can I calculate something like
sink rate using just the GPS fixes ("B" records)?

I'm thinking about taking an early morning tow sometime this winter
when the pressure is high and the air is stable (at least vertically)
so I can calculate my actual polar, but I don't want to waste time and
money on this if the flight log isn't good enough. (And a few cockpit
notes of course -- IAS at various times, etc).