Initially the problem did not look like it was at the API level as suggested by jasfox. But in the course of discussion, it was noticed that there was some problem indeed with the way Chrome fired the unload event whenever mailto: protocol got fired in the window which caused the maps to unload. A work-around was suggested which seemed to sort out the issue as accepted by the original poster.
We selected this question because we thought that it was a good idea to highlight a query which not only exposes certain loopholes, but also provides work-around to get around it.
Keep those interesting questions coming!
See About Featured Discussion Board Posts for more information.