The tips from ballardw are correct.   The problem stems from the US map only going down to the state level, but the MX map going 1 step more granular than state. Here's a modified version of your most recent code, with some "gremove" added to remove the most-granular level of id, and produce a MX map of just the 'state' outlines.   data us_map; length map_id $6.; set mapsgfk.us_states (where=((statecode in('CA' 'AZ' 'NV' 'UT' 'NM' 'CO' 'TX')and density<=3))); map_id=statecode; data mx_map; length map_id $6.; set mapsgfk.mexico (where=((id1 in('MX-02' 'MX-03' 'MX-08' 'MX-26' 'MX-05') and density <=3))); map_id=id1; run; proc gremove data=mx_map out=mx_map; id id; by map_id; run; data usmx_map; set us_map mx_map; run; proc gproject data=usmx_map out=usmx_map latlong eastlong degrees dupok; id map_id; run; proc gmap map=usmx_map data=usmx_map; id map_id; pattern1 value=empty; choro segment / levels=1 nolegend coutline=gray88; run; quit;   ... View more

Read through the explanation in my book (Example 9, starting on p. 63) and see if that gives a better understanding of how everything works.   http://robslink.com/SAS/book/GraphBeyond.pdf   I would recommend getting one thing working separately at a time (the ginside, the annotate, etc), and make sure you know how each piece works, before trying to get it all working together. ... View more

In the "best case scenario" you could overlay the two using proc greplay. But even then, the depth/layer perception would be off, because greplay just overlays them (it doesn't blend them together in a 3d way). Unfortunately, SAS/Graph is probably not the best tool for 3d modeling.   ... View more