I often during a workday need to collect ideas that need to be processed and executed on that same workday. Since it can be problematic to mix these things with other things coming into my inbox that are not as urgent, I have been playing around with using two inboxes instead of just one. The normal one is run down to zero every 24-48 hours as per standard GTD, the "urgent inbox" is processed at the first possible moment. This, at least so far, gives me the possibility to capture potentially important stuff in an inbox without the need to interrupt what I'm working on at the moment for full processing, while at the same time knowing that I'm still fast-tracking into my system whatever new stuff needs to be fast-tracked.
For email of course, since I don't have control over how stuff land in my inbox, a regular emergency scan a couple of times per day is still necessary, but with paper I could do this urgent-or-not processing up front so that urgent input doesn't even mix with regular input.
Physically, my normal inbox is a regular in tray sitting on my desk, with all stuff in it in a nice pile. The "urgent inbox" means sticking stuff to the wall slightly up and to the left side of my field of view when working at my desk, so that I'm not distracted by seeing it while working on other stuff, but also so that I can't ever miss it as soon as I put one thing down to start on the next. If there is more than one thing, they are put next to eachother, so that I do not need to sift through a pile to get at what I need to.
Still, this is something that I am still only starting to try out... If anyone actually has had a system like this running for a while, I'd be very interested in hearing about how it works out in the long run.
As for improving effiency in the slightly longer term, if inbox zero every 24 hours is an issue, at least for me I know that being able to shorten the average time that I need to process an item is what would make the the most difference. If there are 50 items in the inbox, spending on average 30 seconds on an item takes 25 minutes; being able to get down to 10 seconds per item on average will shave that down to just over 8 minutes - which should make quite a difference. I expect this to come partly with experience of working with GTD, and partly by being conscious to the process (I do have a small timer at my desk and use it now and then to get more accurate feedback on how I actually make use of my time, since I am aware that I have problems estimating time realistically.)