Eleazar Zuniga <ezuniga@taoglas.com> says in email of 28 Nov 2011 that we 
should be able to use any of the pin or SMT patches other than the 35mm ones,
as long as we have ground plane under the patch itself that's as large as the
patch or bigger.  I'm not sure I believe that, but adding an 0402 site to 
ground for a tuning cap ala what the chip antenna docs call for might be a
reasonable thing to try?

GP.1575.12.4.A.02
	12x12x4 mm, pin plus adhesive
	gain towards zenith -0.5 dBic

GP.1575.15.4.B.02
	15x15x4 mm, pin plus adhesive
	1.0 dBic at zenith

GP.1575.18.2.A.02
	18x18x2 mm, pin plus adhesive
	gain towards zenith -0.5 dBic

GP.1575.18.4.A.02
	18x18x4 mm, pin plus adhesive
	gain towards zenith +3.5 dBic
	$4.13 at q100				*** 50% the mass of 25mm

GP.1575.25.2.A.02
	25x25x2 mm, pin plus adhesive
	+2.0 dBic gain at zenith

GP.1575.25.4.A.02
	25x25x4 mm, pin plus adhesive
	gain towards zenith +3.5 dBic
	$4.05 at q100

surface-mount versions

SGP.1575.12.4.A.02
	12x12x4.5 mm
	-2.0 dBic at zenith, 4dB axial ratio, 6Mhz bandwidth

SGP.1575.15.4.A.02
	15x15x4.5 mm
	+1.0 dBic at zenith, 3dB axial ratio, 6Mhz bandwidth

SGP.1575.18.4.C.02
	18x18x4.5 mm
	+1.0 dBic at zenith, 4dB axial ratio, 5Mhz bandwidth
	$4.94 at q100

SGP.1575.25.4.C.02
	25x25x4.5 mm
	+2.0 dBic at zenith, 3dB axial ratio, 5Mhz bandwidth


From all of this, GP.1575.18.4.A.02 would seem to be the clear winner.  Just
over 50% the mass of the 25mm version, same gain, only a few pennies more...
and much cheaper than any of the SMT versions with better specs.  I also think
that soldering in the pin is likely to be a great mechanical retention 
solution.