Sorry for the long answer, but I have a long quotation that's quite on target and I'm trying to include all the facts that I found that pertain to your question.
I'm going to my source for this, one I've cited here before, the Star Trek: The Next Generation Writers' Technical Manual, Fourth Season Edition. This was sent to me directly, in a pitch package by the ST:TNG Script Supervisor Lolita Fatjo, when I was invited in to pitch story ideas to the producers. This is one of the Writers' Guides sent to those who would be writing for the show and, in some ways, goes beyond canon because it defines canon. In other words, it tells the writers what they can and cannot do on screen.
On page 28, under The Transporter - Once and for All:
... The stream of molecules read by the pads is sent to the Pattern Buffer, a large cylindrical tank surrounded by superconducting electromagnetic coils. It is here that the object to be transported is stored momentarily before actual beaming away from the ship (or even within the ship). It is the Pattern Buffer and its associated subsystems that have been improved the most in the last half-century. While the actual molecules of an object are held in a spinning magnetic suspension (eight minutes before degradation), the construction sequence of the object can be read, recorded in computer memory (in some cases), and reproduced. There are limits to the complexity of the object, however, and this is where the potential "miracle" machine still eludes.
The Transporter cannot produce working duplicate copies of living tissue or organ systems.
The reason for this is that routine transport involves handling the incredibly vast amount of information required to "disassemble" and :reassemble" a human being or other life form. To transport something, the system must scan, process, and transmit this pattern information. This is analogous to a television, which serves as a conduit to the vast amount of visual information in a normal television transmission.
(Note: This was written when all TV was still analog and DVDs were not yet out -- and remember that we could transmit TV signals before we could actually tape and keep them.)
Continuing to quote from the same passage:
Storing that information, however, is another matter. In our analogy, it would be like comparing a television (which is incapable of storing an image) to a videocassette recorder, which can store a relatively low-resolution recording of a television program. In order to store the patterns for a human being, one would have to record not only all the atomic and molecular configurations, but all the quantum and energy states of all the electron shells, and the brownian motions of every sub-atomic partical of every atom. While we cannot store all of this incredibly complex information, we can use it as it is being handled in real time.
I've seen this with computer data in some programs I've had to write. Data came in so cast I could filter and pull what I needed, but I couldn't store it all (without filling my hard drive up).
In other words, the pattern buffers store actual atoms and molecules during transport and the information relating to them is more than can be stored in a computer. Storing the data for a cup that doesn't depend on things like Brownian motion requires much less info. Just store the data on the materials and the shape it needs to be in, and when you replicate it, just build the shape out of the needed molecules.
I think this includes most of the answer. There were cases, notably in Wolf in the Fold in the original Star Trek, where matter (in this case a human body), was beamed out into space on a wide dispersal pattern. I've never seen an explanation why the buffer just can't be purged.
I would assume (see, I'm on dangerous ground here, but I'll admit it!) that the reason they can't just purge the buffer is because, as is pointed out above, the buffer does not contain data, but actual particles, so they can't just dump them or clear them out like they would if it were a body (as they did in Wolf in the Fold). They'd be dumping the particles somewhere, but they'd still be Omega particles.