You get a big thumbs down for making it impossible to report a bug on google spreadsheets. No, I don't want to post to a forum. Set up Bugzilla, and stop wasting 15 minutes of my (and presumably others) time bouncing all over your site looking for a way to submit a bug report.
P.S. Your array evaluation mode is busted. If it's not powerful enough to do something, failing is better than doing something arbitrary. I admittedly am not a spreadsheet guru (although I have reluctantly learned a few tricks from some).
I don't use Excel outside of very rare circumstances (I don't have a copy except on an old battered system - it's almost likable for a spreadsheet), I use GNUMeric a fair bit (very fast but a bit buggy), and Openoffice is useful to translate things between other formats (but is too heavyweight for my tastes). Google spreadsheets is a young product but looks to be slowly reaching the "reasonably useful" state. It seems like spreadsheets have the same kind of loose communal standards that SQL databases do - just irritatingly different that migration between them doesn't work very well, a few small differences, and some areas of advanced functionality where some of them are so different or deficient that one has to do a lot to replicate what's easy and done in another.
Rough areas of SQL database 「standards」 tend to be related to Relational Calculus, performance issues, constraints/triggers, data types, and cursors. Rough areas of spreadsheets tend to be in different evaluation styles (array evaluation, solver functions, etc), lookup/matrix functions, and data types. In theory, SQL is standard. I don't know of any standard for spreadsheets. Both are in about the same level of mess. The benefit to the standard is that we can blame Oracle or IBM or Microsoft or the Postgres devs when they don't adhere to it (and they can then say "yes, we don't, maybe we'll fix it if we feel like it"), while Microsoft, SUN (openoffice), or whomever presumably started out by copying each other and any interest in compatibility came from their early years of making migration to their product easy.
If I had a mac, I'd probably like trying Apple Numbers - I keep hoping to find something like Lotus Improv. I suspect for most of us who stay around software systems for long enough, we outlive a number of ways of doing things and carry that memory of something better with us for awhile (even if that "something better" is later surpassed). I'm sure there are people still around who take things a bit far and cling to their ancient Amoebas, cubes, or sufficiently old PCs with Matrox video cards for OS/2.