The nominal impedance of a xfmr is an abstraction.
The only thing it describes with a relative dose of certainty is that when the secondary is presented with the nominal load, the primary will measure as the nominal primary impedance, within the limits of rated frequency response.
In the particular case of your 150:15k xfmr, if you connect it in reverse, loading what is now the secondary with 150r will result in 15k at the primary.
Now, you would not connect it to a 150r load, as you would connect it to the input of a mic preamp.
According to usage, the actual impedance of a mic preamp rated for nominal 150-200r microphones is supposed to be about 10 times higher, for 1500/-000r.
The 15k:150 xfmr, presented to this 1.5-2k load, should reflect at the primary as 150-200k, if the xfmr was perfect.
Unfortunately it is not.
Particularly at low frequencies, where the inductance of windings become dominant, but also at high frequencies where capacitive losses intervene, putting an inferior limit to the actual impedance.
So the impedance seen by the instrument may well be about 150k at midrange, but much smaller at LF and HF. The result is a midrangey frequency response.
Reiterating the calculations with your 50k:250 xfmr shows that there is a predictable advantage. However, since the measurement conditions that produced data are probably not the same, it is impossible to predict the actual result.
Many passive DI boxes use a mic xfmr in reverse; results are variable, but the most significant variable is the user's expectations.
Personally, for years I've played a passive-piezo equipped guitar with a passive DI, which in theory should result in something unusable, however with the help of some EQ it gave a satisfying result.
Now all attempts I had with playing electric guitar through a passive DI resulted in nothing good, when active DI's produced a usable result.
However, I maintain that for electric guitar, there's nothing better than moving air via a piece of carboard, although some amp simulation softwares give pretty impressive results.
The latter is the justification for an active DI input, not passive.