[jccooper]

Mostly, I think, because they'd prefer to avoid having several types of engine involved. SpaceX is big on cost savings through commonality. The development and tooling and testing etc. for creating a smaller landing-only engine would probably cost darn near as much as for the main engines, so if you can get it done with them, it makes sense to do so.

If they can't get them on the ground reliably with just the Merlins after trying enough, then maybe they'll do something like that. But if it's a simple matter of software and a handful of tests, as it seems may be the case, then they've saved a lot of effort and cost.

The other answer is that the F9 architecture was designed (and flying) before this mode of recovery was "the one". And they'd prefer to not redesign the thing if it can work as-is. So they'll try it and see before spending gobs of money on a re-design. Rockets aren't software (except to the extent they are) and changing things about them is dangerous and expensive.