Okay, it appears blockcypher count the sats more accurately.
Notice that some transaction has a "more" function which brings this if you expand:
"Estimated Value Sent" excludes known change addresses. For example, let's say we have a single transaction where address A sends 1 BTC to address B and also 1 BTC back to address A as change, then only 1 BTC is estimated to have been sent. Proper use of a new change address for each transaction (like all HD wallet implementations) obfuscate this feature.
If we take the same situation in mempool.space: When 1 BTC goes back to address A as change, it will still be counted as a new addition in "Total Received" which inflates the amount.
Take a look at address
bc1qhzdcqgeta3kh5an7354zln7ghmznlrm0ltnher with far less transactions made.
Blockcypher total received: .12
Mempool.space total received: 0.38284160
After receiving .12
BTC, succeeding transactions used the same change address as the input's 'every time' so it's quite obvious this address has only received .12
BTC. Now, count the sats everytime address ...m0ltnher is an output in a transaction and it'll equate to 0.38284160
BTC