Users should have to play the complete works of Sorabji to go from Basic to Standard, and the studies of Ligeti to go from Standard to Pro - sending you AVIs to prove it was done in real time.
Just kidding. I would go with taking the difference and adding 30% or thereabouts - that way anyone who thinks there's the least chance they will be quickly persuaded to update might just go straight for the next version up to start with.
The difference plus 30% makes the upgrade path a bit steep in my opinion. Maybe the difference + 10% - 20%? That would still make the more expensive versions interesting for those who can afford them, while at the same time leaving the less fortunate with a decent upgrade path.
Or the difference + 10% - 20% of the price of the-upgrade-to-version. That would make a one-step upgrade (from basic to pro) cheaper than a two-step upgrade (basic-standard-pro), thus maybe keeping the number of upgrades down.
I second the idea about 10-20% overhead on the upgrades. 30% seems to be bit too much, yet no overhead would - as Gary predicts - cause the avalanche of upgrades.
However, the upgrade from downloadable basic version to anything higher could be with less overhead, as download needs no (am I wrong?) processing of Garritan people whatsoever. I assume it's completely automated, or will be - if is not yet available. If that's the fact, for Gary the upgrade would be the same effort as first buy then.
p.s. the samples of Steinway just blew me away. Especially scherzo, where the "meat" in bass notes is easily heard, and is not mudded like in other ones.