There is a low limit to the maximum size of an SMS message. Small enough to fit the entire thing plus all meta data easily into a 512 byte UDP packet. Sure, UDP is unreliable, but everyone knows that and you build additional reliability mechanisms on top of that. The SMS protocol is different, but still shares a lot of similarities.

With iMessage, the messages themselves can be much larger, and the larger the message the more likely it may run into packet fragmentation which could easily result in it being silently lost down the road.

With SMS, once the network has accepted the message, I think you have a higher probability that it will actually reach the recipient.

Another factor here is cost — precisely because you pay for SMS messages, the carriers are going to aggressively maximize the delivery probability, because they don’t want to lose a single red nanocent. But iMessage is free, so why should anyone care to optimize that for maximum delivery probability?

Then you also run into the other issues noted above, e.g., where someone previously used a given number with an iPhone and then moved to Android. The number is still detected as being iMessage capable, but you can’t actually send an iMessage to them on Android.

I guess all I’m really saying is TANSTAAFL — There Ain’t No Such Thing As A Free Lunch.

Oh, and be careful what you optimize for. You may optimize yourself out of a job.