Is the make table to access local?
I mean a dog and say a bus are two VERY different things here.
Is that make table local to oracle, or are you taking about the PT query being used as a make table to a LOCAL table in Access? As such, your comparison is a mount Everest of a difference here.
I suspect that a plane jane linked table to Oracle vs that of PT query likely would perform the same if the target table is "local". That's because you pulling the whole table down the network pipe to a local table.
If the PT query is a make table 100% to another table on the Oracle server? Then a PT query should perform just as fast as if you typed that command in the Oracle console - there should be no difference.
You don't mention the row count
(EDIT - i see you note only 16,000 rows - that is rather small).
So the issue then is the network connection. is this a LAN (local area network), or are you using say a VPN and the internet which will be about 100 times slower. I mean, a connection that is 100x slower then a typical LAN is going to be 100 times slower - no surprise here.
So, details as to the type of connection being used, and the row count would shed significant light here. But if a VPN and connection over the internet is involved here, then such connections are about 50 times slower then your super cheap-o office local area network.
You can read this article of mine that explains this speed issue:
http://www.kallal.ca//Wan/Wans.html
I don't see why 16,000 rows would take so long - that will take well under a second of time on the cheapest office network. Using any other client to pull that data should not run or work any faster then a PT query. I mean, ado.net, ODBC, or oleDB don't cause the server to say, hum, gee - lets send the data slower. All of these clients should pull data at the same rate.