Ideed you are absolutely correct, I had points in my LatLon list with
invalid coordinates due to a mistake in my NMEA decoding
function. After correcting this mistake and getting all the points
in the right range, I can indeed observe the acceleration associated
with the use of the Encoded Polyline and most significantly the
fact that I no longer have to cut my trace in small 100-point long
GPolyline but I can now display the whole trace as a single, big,
encoded polyline. I can easily imagine now that due to these erroneous
points, your algorithm for defining the level of each point was
confused and
assumed it had to plot all the points at all zoomlevels.

Indeed my objective was to write a script for anyone to put an NMEA
file on the web and point the script to the track file for displaying
over
GM, hence the need for dynamic encoding of the polyline.

Thank you for your prompt help, Jean-Michel