That's a valid point.

In theory there would be ways do deal with this directly, but the solutions occurring to me (preserving the original concept of no fee-competition in the future block) would make the protocol much more complex and also less useful. For example, one could think about limiting the reservation to a timeframe (between blockheight X and Y), limit the amount of "reserve transactions" per block, etc..

But a relatively simple solution could be to change the concept a bit:

Instead that the "reservation" means that you will be able to get a space in any future block without fee competition, it means that you are enabled to compete in an additional space of the block (e.g. 4 MB vBytes more). In this space there will be less fee competition so a fast confirmation is extremely likely, but not guaranteed. This would solve the "huge block" problem and also gives the miners even more incentives to include the second transactions, as there will be to some extent a fee market in this additional block space because it's not unlimited.