Hi all,

attached is a patch that addresses the problem described be Roger Calvert:
http://gis.19327.n5.nabble.com/Incorrect-compilation-of-grid-lines-tp5809502.html

The problem is caused by the fact that we use a 2D formula for objects which are on
a spherical surface. In most cases the error is very small and can be ignored, but
sometimes it isn't. We use the formula to decide whether a point is important or
not, so I decided that to keep all points where the error is big.
A better way would be to use a correct formula in that cases, but I am not clever
enough to code the correct formula  in Java :-(

@Roger: Please check if the results are ok for you.
BTW: osmconvert seems to have a problem with some coordinates, e.g.
<node id='90053000500' lat='54.3788936' lon='1.67E-05' version='1'/>
is interpreted as
<node id="90053000500" lat="54.3788936" lon="1.67" version="1"/>
Don't know if the OSM api allows the exponent format?

A compiled patched binary for mkgmap can be found here:
http://files.mkgmap.org.uk/download/217/mkgmap.jar

Gerd