1. To talk clearly about download speed being "too low" we need to establish what the correct
expected download speed should be. That involves such
limiting factors as DSL sync rate, which may be lower than DSL profile, as well as ATM protocol overhead.
2. It is essential to use
accurate measuring instruments. Some tests that have been recommended in this forum frequently show impossible results and cannot be trusted. (A test that reports better results is not a better test.

)
3. It is necessary to know
how to use the measuring tools. Other traffic on one's system or network will result in lower reading. As can QoS on the router, for example. Or a busy PC. If a test does not produce a sequence of
consistent results, either the tool is defective, or your network/system is too unstable to be measured (and expressed by a static value).
4. It is essential to
distinguish clearly between
network speed and
application performance. When we say "download speed", are we talking about our network data throughput capacity, or are we talking about application performance, which may report "speeds" that may be affected by application delays, compression, inefficiencies, application interdependencies, PC constraints, or errors.
5. It is important to check measurement units, overhead values, and arithmetic, whenever mentioning numeric values.
In particular:
ATM protocol encapsulation overhead leaves 85-87%. Thus, a sync rate of 5056 kb/s cannot exceed 5056 kb/s *.87 = 4400 kb/s (550 kB/s) theoretical maximum throughput. 85% would be a more conservative limit. Using this limit we get:
Sync-Rate Max-Throughput(85%)
3008 kb/s 2557 kb/s 320 kB/s
4032 kb/s 3427 kb/s 428 kB/s
5056 kb/s 4298 kb/s 537 kB/s6016 kb/s 5114 kb/s 639 kB/s
7040 kb/s 5984 kb/s 748 kB/s
These are "theoretical" upper limits, assuming that the data is deliberately constructed for maximum throughput. A speed-test can arrange its data in that manner -- making each packet long enough to just fill an MTU, and to fill all the ATM cells. Real-life data packets are not optimized for speed-test results.
Average data encapsulation is about 80% efficient, rather than 85-87%.
A realistic throughput limit in Bytes/s is: 10% of the sync rate in bits/s.
(It would take a sync rate of about 10 Mb/s to achieve a throughput of 1 MB/s.)
The answer to the OP's question is:
yes, with 5056 kb/s, the maximum download throughput is about 500 kB/s.
The above calculations are based on the following values (sizes in bytes):
53 ATM cell size (fixed)
5 ATM header
1500 ethernet frame limit (MTU)
8 PPPoE header
20 IP header
20 TCP header