That's the right way to test the transformers be they the 1:9 or the higher 1:49 or 1:64.
Most I've tried maybe a DB of loss. Not much at all.
Most matching networks are as good be it an L network or T.
The problem is with the 1:9 and the random length of nonresonant wire is that the impedance at the feed point
can be all over the map. If it's too reactive or low in impedance the transformer takes a beating due to the
high reactive currents. Can't blame the transformer for that. There are two magic lengths 53ft and 124.5ft
that produce the least SWR usually under 2:1 and tend to not beat up the transformer. With a good counterpoise
or as a inverted L with ground grid under them they perform well.
Random length wires that are shorter can be hard to tune be it a transformer or transmatch if it is the wrong length. Mostly due to
high capacitive reactance.