TrigpointingUK

The manual for my Garmin GPS device states: "your GPS location is accurate to
within +/- 12 ft." This may be fine for many purposes but when searching for
(say) a Buried Block, thats a circular search area of about 450 square ft which
isn't feasible without further clues.

I'm wondering if the accuracy improves when measuring a fairly small relative
distance rather that working with absolute grid references. In other words,
does the error term remain roughly constant for a given locality and GPS device?

A case in point:
I have an OS sketch which specifies a distance of 29.95 metres from the
NE corner of a building to the block. A second measurement gives 26.52 metres
from the SE corner of the same building to the block. Other measures on the
sketch are to reference points that no longer exist. But the intersection of
two circular arcs, with radii of the two distances, would give the required
position.

Would the avoidance of measuring an absolute position be likely to give a more
reliable result than simply entering the grid reference into the GPS and then
moving to the supposed location?

Or would the use of a low-tech tape measure be a better strategy?

Measuring tape, every time.

Assuming we're working in BNG, variation from the true WSG84 position alters steadily, depending exactly whereabouts on these islands you happen to be standing; can't remember why, probably something to do with BNG's failure to incorporate any ellipsoidal calculation. But more significant, because incalculable, the same handheld device will also record different readings on the same spot on different days owing to atmospheric conditions or whatever. Even with a state of the art thousands-of-pounds-worth of Leica you could never replicate the precision of a measuring tape showing millimetres.