> "What is known with a fair degree of certainty is the
> time between proto-Indo-Aryan and the modern Indo-Aryan languages --
> something in the order of 1,000 years.
Dixon is, alas, wrong, or else you have inadvertently typed the wrong
number. We have Sanskrit attested from c700BC, and other evidence suggests
that some of what is recorded by 700 BC was originally written 400 or 500
years earlier than that. So the known time from Old Indian to the modern
Indo-Aryan languages is at least 3000 years.
This, however, is a sideline to your main argument. The simple answer is
that, as in so many branches of human enquiry, we can do no better than
build a "best explanation". Of course alternative dates are possible, but
they are not all as likely! There are a number of methods we can used to
give an approximate date for a proto-language.
(a) Despite your understandable caution, we do have a rough idea - very
rough, but still helpful - of how much a language can be expected to change
in 3000 years compared to 500 or 10 000 years. The changes necessary from
PIE to the proto-language of each family within IE can realistically occur
within 3000 years; it would be much less believable to suggest they
happened in just 500, or that they took 10 000.
(b) We can look for loanwords in that proto-language from other known
language groups. So Latin and Greek are particularly important for dating
protoGermanic. In the case of PIE, we have Semitic loanwords in PIE, such
as the numerals 6 and 7, and other words. Since the loan form is
reconstructed for PIE, not just a few branches of it, PIE speakers must have
been in contact with a Semitic group. The Semitic languages are attested
1500 years earlier than any IE language.
(c) placenames give clues. When a place has a IE or PIE name, we can guess
that at some stage PIE speakers were in that region. This gives us links to
evidence from archaeology. Although we cannot know what language a
non-literate culture spoke, we can build a picture of movements of peoples.
This evidence is particularly important in tracing and dating the movements
of proto-Indo-Iranian peoples, and gives fairly precise dates for what is
possible. So we get good approximations for proto-II, which helps refine
our projected dates for PIE.
(d) This post is already too long - go and read one of the standard books,
which can explain all this much better than I can, with good examples.
Peter