Material studied shortly before sleep has the consolidation window immediately available, so it is replayed during the slow-wave-rich early night. Material studied in the morning competes with a full day of new encoding before the next sleep opportunity arrives.
Two students, one exam
A student has an exam in two days and 90 minutes of material to learn. Student A studies from 9 to 10:30 p.m. and goes to bed at 11 p.m. Student B studies from 7 to 8:30 a.m., then attends a full day of classes before sleeping. Both spend the same 90 minutes. Student A's material enters the first slow-wave-rich part of the night within an hour of encoding. Student B's material must survive a day of interference before the next sleep window, and if Student B also cuts sleep to study more, the second half of the night — where REM is concentrated — is the part lost.
The trade-off that matters
Trading sleep for extra study time removes the very process that stabilizes what was studied. The prediction is not that late-night study is always better, but that the sleep interval following study is part of the learning, not a pause from it.