This has nothing to do with tracker site configuration settings although this may be loosely caused by tracker site "settings" in the sense of location/events of a story in a book.
Some peers even on private trackers may be running on ADSL or low-end cable lines as well, they'll get "firehosed" even by a single 100 mbit/sec upload seed. You definitely won't be able to upload quickly to them when that happens.
Even a fast line may be so badly overloaded by other activities (other torrents for instance) that their "spare" bandwidth is tiny.
A downloading peer might be getting the bulk of the torrent from lower-latency and lower packet loss seeds even though those seeds are not the fastest ones available or connected ...possibly due to: https://en.wikipedia.org/wiki/Bufferbloat
So there's a couple reasons why seeds and peers can appear to be a magnitude (10x) faster or slower than each other.
The only way to "balance" the upload speed equally between a bunch of random peers of unknown quality is to run everything at the speed of the SLOWEST peer ...which doesn't sound like a good idea to me.
While downloading, the BitTorrent protocol's Tit-For-Tat logic is supposed to give MORE upload speed to peers that are uploading to it.
While seeding, more upload speed is supposed to be sent to the peers that can download the fastest. These are the peers under the least load on the fastest line with the lowest latency and packet loss...and possibly not uploading to any other peers at all!