I think I get it. 
So you basically extract all the transactions from the blocks and store them in a separate files (transactions1+transactions2), giving each transaction a unique sequencional "numeric index".
Then you have the index file (tx-index) where each different TXID points to the "numeric index" - that's actually your equivalent of UTXO-index, except that it includes also spent transactions.
And then you have the spend-index (addressed by the "numeric index") which basically tells you whether a specific output of a specific transaction has been spent.
Is that correct?
So you basically extract all the transactions from the blocks and store them in a separate files (transactions1+transactions2), giving each transaction a unique sequencional "numeric index".
Then you have the index file (tx-index) where each different TXID points to the "numeric index" - that's actually your equivalent of UTXO-index, except that it includes also spent transactions.
And then you have the spend-index (addressed by the "numeric index") which basically tells you whether a specific output of a specific transaction has been spent.
Is that correct?