Of course his views (and most everyone elses) evolved when Lorentz derived the Lorentz transform on grounds of a physical argument, and pointed out that Maxwell's Equations were maintained their form after such a coordinate transform, in gravity-free universes (simple linearly related inertial systems).
When Einstein developed his special theory of relativity, he rejected the putative measurability of the ether in flat spacetime metrics.
Because the ether does exist, when considering the wider world of non-flat spacetime metrics.
You could say Michelson-Morley compared 2 horizontal axes (which are rotation symmetric with the gravity vector). Variations of predictable ether effects are perfectly measurable, when comparing horizontal with vertical paths!