Thoughts about GPS and elevation

I've been using my iPhone to record all of my rides for a year now. One thing I find interesting is how big the margin of error is. The official mapmyride calculation for ascent http://www.mapmyride.com/routes/fullscreen/52396398/ says 2877 feet of gain. The actual gps record of my ride on sunday
http://www.mapmyride.com/routes/fullscreen/52892664/ says 2894 feet of gain.

The cyclemeter summary and split says 5897 feet of ascent (http://maps.google.com/?q=http://share.abvio.com/bc64/6357/4c7a/cf0e/Cyclemeter-Cycle-20111002-0743.kml).

Finished Cycle: Oct 2, 2011 3:30:44 PM
Route: 2011 Major Taylor Century
Google Maps URL: http://maps.google.com/?q=http://share.abvio.com/bc64/6357/4c7a/cf0e/Cyclemeter-Cycle-20111002-0743.kml
Shortened Google Maps URL: http://j.mp/ogumTC
Import URL: http://share.abvio.com/bc64/6357/4c7a/cf0e/Cyclemeter-Cycle-20111002-0743.kml
Ride Time: 6:32:23
Stopped Time: 1:14:52
Distance: 100.67 miles
Average: 15.39 miles/h
Fastest Speed: 47.63 miles/h
Ascent: 5897 feet
Calories: 4386

The thing is, when I compare the actual elevation from the phone to mapmyride, it is off by 100 feet. In some spots it's off by 130 feet. This is unscientific, but it appears weather plays a big factor in the precision. On overcast and cloudy days, the elevation has a higher margin of error.