This example doesn't make any sense, as you said...
In the first case you are loading four months, then you are loading the same four months into the same table (two tables with the exact same fields are automatically concatenated by Qlik), then you are loading the ensuing eight lines and adding a field. The first Load * doesn't even do anything since you're doing a preceding Load * from the same Load *. MESES2 is not created because the first set of loads was automatically concatenated to MESES, and the second one is implicitly concatenated to MESES because it's the previous table.
In the second case, you're using a preceding Load * on a version of MESES to which you've already concatenated the extra field, so you're getting four lines total, matching the original number. Because the list of fields in this table isn't identical to MESES, the tables aren't automatically concatenated and MESES2 is created.
I though that when using "preceding Load" in Qlik it works from bottom to top, so in my example, shouldn't it execute the code from the "conatenate" keyword and obtain 8 lines, and then add the "LOAD *;" at the top? And shouldn't the LOAD * just load all the data behind it, (8 lines)?