OK, to appreciate the problem, you have to know how much copper wire there is between your modem and the nearest broadband node. The farther the distance, the poorer your performance will be. Secondly, you must realise you are sharing your bandwidth with everyone else; there is a limit. To explain this, let me try to give you a purely hypothetical picture:
Imagine you have a link in the chain that has a limit of, say, 20 Mbit/s and that you have 100 subscribers using that link. At 3.30 am, you may be the only person on that link, and you are very close to the node, in which case you will benefit from the full performance you pay for, say 2 Mbit/s down ½ Mbit/s up. By 0630, 9 other similar subscribers are on line. The ten can share the 20 Mbit/s that is available and still get the full performance (actually slightly lower, but let's ignore the abstruse technicalities). By 0830, 30 similar subscribers are busy downloading something. Because they are sharing the bandwidth, each has only 20/30th = 0.67 Mbit/s.
Now, going back to my first point, copper pairs, used between the fibre optic node and your house use the upper frequencies for ADSL. The impedance of the pairs cannot be matched to the terminal impedance of the splitter in your phone socket, so you get reflections along the line and these reduce the reliability of your communications. The longer the wire or the poorer the quality of your wire and the equipment at both ends of the wire slows down the bitrate until the handshakes confirm that what you down/upload are going through reliably. With ADSL, the maximum theoretical length of copper in a connection is about 5 km. In practice, this may be as low as 3 km and the nearer you are to this limit, the relative up/download rates will approach them.
I am exactly 1.7 km from the node to which I am connected. I pay for 2 Mbit/s download. At a favourable time (few others using the node), I get about 1.3 Mbit/s download, rarely more. I surmise this is the limit imposed by the copper for my type of subscription. At busy times of day, my download speed may drop to 600-800 kbit/s, because the node is being shared with other users.
This is the way ADSL works and there is not much you can do about it.
However, you should never forget that the "throttling" of download speeds is not just the prerogative of Cytanet. A typical Internet connection may go through anything between 10 and 30 nodes and any one of these can slow the transfer of data if the traffic is high, not to mention that the server at the other end may suffer the same problems as you have, especially if it is connected by copper (most decent ISPs have fibre optic, but not all), and the traffic to that server may be high at a given moment. The data that is transferred between the two ends of the connection is sent in "bundles" (called packets) and each packet is checked by a handshake before the next one is sent. If it fails the check, this is repeated until the handshake says that the transfer is OK.
Take cyprus.angloinfo.com (this forum) as an example. The server is in San Antonio, Texas managed by ServerBeach. It takes me through 23 hops, 7 of which are in Cyprus, to reach them. It takes 28 ms with no lost packets to get me out of Cyprus, which is good. From there, it goes to a node in London, run by BeyondtheNetworkAmerica and this node slows me down to 118 ms, which is less good but acceptable. What happens after that, goodness only knows, because I get 100% data loss at the intermediate nodes most of the time, and this is why Angloinfo is excruciatingly slow at times. This is nothing to do with Cytanet and everything to do with Angloinfo and the path that their slow server at ServerBeach takes to get to London.
In contrast, one of the servers I use for cypenv.info is in Cincinnati and it takes 22 hops to get there. Most of the hops give 0 or only 10% data loss for a series of 10 packets (10% means that one of the 10 packets has to be resent, but it gets there) and the return journey takes only an average of 220 ms, which is quite usual for a transatlantic connection. Another server I use, at cypenv.org, is is Pittsburgh and is reached in only 12 hops with a return time of 199 ms, which is very good for a transatlantic connection. However, this one is a bit of a cheat because they use a semi-private link between London and Chicago and they have their own private 10 Gbit/s fibre optics between Chicago and Pittsburgh.
The point I'm making is that slow downloading is not always Cytanet's fault, if it is at the other end or anywhere in between. Even the node that Cytanet connects to (there are half-a-dozen different ones) is usually slower than Cytanet itself. And the speed-meters do not give you a true indication, either, as they depend on the time a packet is sent somewhere and then comes back. The answer you get depends as much on the "somewhere" as on Cytanet.