But looking at average TX fee rate in last few months, personally i'd avoid creating UTXO with less than 10K satoshi.
True, but:
1. This topic is about "magic numbers". Note that on my list, there are exactly 21 possible values. I don't want to make it 18. And I don't want to start for example with 500 BTC, to end up with 10K satoshi. It is hard to make things tick, so I will leave it as it is.
2. You don't have to use lower coins, if you don't want to (the same with the upper coins: for most people, including me, this list starts with more coins than you own, and that is done on purpose). It is hard to find a wallet in the wild, that will create round numbers, because it is even a very strong suggestion, that you should not use such amounts (and your privacy score is then decreased on many sites, if you stick to some list like mine). Which means, you have to use coin control, and pick everything manually anyway, no matter what fee levels are there (and, guess what, using round coins, and round fees, is even harder, so usually in the current situation, you can for example put your last four digits into fees, if you want to use 10K as a lower bound).
at the end if we need to spend some round input there should not be a problem
Because that is the point: if you have round numbers, it is harder to spend. And I can see some positive sides in this, because if I would use non-round numbers, then I guess today I would have 10x less BTC, if not less. Round numbers are useful to remind you: "protect your privacy, don't touch those coins". Because if you are somewhat exposed by using round amounts, then you need some extra effort to move those coins in the future. And then, you can easier resist the urge to buy or sell your coins, because the round amount keeps you from moving it at all, unless you know, how exactly you want to mix it, and when you are ready to take the full effort of doing that properly.
So, what are you gonna do? Save or spend? Choose wisely, and pick your side, by using proper amounts.
This is not the first time, when I disagree with garlonicon: I want to save coins, he wants to spend them. I want to dig deeper into hash functions, he thinks there was no hashing in some challenges, and focus on elliptic curves alone. But yes, the general advice is good: clearly choose your goals, and then adjust all amounts, accordingly to your main goals.
Another thing is that this particular list was created with some "one fits all" style. Which means, I want to use the same settings in networks, where one satoshi per virtual byte is always accepted, no matter what. Also, the lower bound of exactly one thousand, makes it easier to port into other networks, for example, in this way, inside LN this list ends with a single satoshi (because of mainchain compatibility), and starts with 0.05 BTC (because test networks can clearly show us, that building a channel with a higher amount is pointless, and can lead to lose of money, so 50 BTC is definitely too high; also it is better adjusted to LN-based mining, where it is very unlikely, that someone will ever mine more than 0.05 BTC, because the whole reward is based just on discounts for your lightning payments, so I doubt anyone will reach a channel partner, that will allow routing more than 0.05 BTC worth of traffic, without getting paid, and will accept mining shares as a reward, instead of taking routing fees).
So, the scale that starts with 50 BTC, and ends with exactly 1000 satoshis, and contains exactly 21 possible values, seems just like a puzzle, fitting the most cases I could imagine. Having exactly 1K satoshis at the bottom, makes it easy to adjust to millisatoshis, microsatoshis, nanosatoshis, and so on, because then you just always move everything by three decimal places, and everything fits perfectly. Also, then your upper bound can show you, what is the maximum amount you should put on such ideas. Which means, that for example LN channel with millisatoshis should not exceed 0.05 BTC. And with microsatoshis, instead of millisatoshis, should not exceed 0.00005 BTC. And with nanosatoshis, should not exceed 0.00000005 BTC. Which means, you should never go further than nanosatoshis, because then you lose mainchain compatibility, if you don't introduce fractional satoshis on-chain. Which is good, because it protects you from activating things like infinitely-divisible coins on-chain (this is bad idea, because then for example Ordinals could move from witness to amounts directly, and that would cause a lot of problems, if you would use "almost 4 MB" to just process some single "amount" field, produced by a pool accepting non-standard transactions).