I don't believe that McClenehan gives a specific reason for why 25,000,
but as stated on p. 377 (pg numbers may be off, I have older version)
"The development of 'K' required a Bayesian approach. However in the
time of Whitney, and for many years to come, the theory underlying
mathematical statistics refused to admit that element of Bayesian
analysis which required a subjective approach to the assignment of a
priori probabilities.... Thus it was to come to pass that casualty
actuaries were to 'sell their birthright for a mess of pottage.' They
'copped out' on calculating 'K' on the basis required by Whitney's
mathematics and settled for an arbitrary assignment of a value chosen by
means most convenient. Having yielded to the demands of the then
popular approach of statisticians, they capitulated even further by
yielding to the demand by insurance buyers and insurance marketing
people for 'full' credibility."
Paul Wilbert