I think variance settings are not properly optimized as when applied, it completely slows down the torrent speed  to very low speeds - closer to 15 KBps.

This does not happen in BiglyBT or qBittorrent or even I2PSnark.

I checked with length  = 1 and variance = 1. And I was not able to cross 60 KBps.

But when I use length  = 3 and variance = 0, I am easily able to cross 250 KBps.

This does not make any sense.

As more nodes are being traveled when length is 3 and not when length is 1.

So, how come we got faster speeds when distance traveled is more??

I think there is a major issue with the way Tixati talks with I2CP protocol of I2Pd daemon and subsequent settings parameters. Maybe check BiglyBT on how it does that for reference.

I am sure you relied on default mode of operation, and node made random tunnels with semi-random peers. Random peers have random speeds, and quite a lot of I2P nodes have low bandwidth limits (see stats site). You can check peer speed limit in its profile in Java I2P console or i2pd config directory. To not to compare apples to oranges, you have to build tunnels manually through the same chain on nodes, then reason about whether their throughput is right.

Moreover, node at the end of the tunnel can be specifically chosen from the ones in far away countries, and long range traffic (that single direct connection) is usually more congested.

Moreover, what's the point of using I2P if you do not care about anonymity? You do understand that one hop tunnel does not provide any magical security?

So, yes, it is normal for I2P tunnels to have very different speeds. It is not made to get you the fastest downloads.