PIE Stop System

Lately I've been giving thought to the PIE stop system and I've found
the proposals I've seen rather unsatisfactory. The Traditional p, b,
bh, etc. system is the most in agreement with the comparative data,
but it's typologically unacceptable. We could add aspirated unvoiced
stops as Szemerenyi proposes but the evidence for these is lacking
outside of the Indo-Aryan branch. At best there are isolated examples
in other families. These stops would also be extremely rare, and
mostly exist in onomatopeia and possible stop + laryngeal situations.
The Glottalic theory at first makes sense but the change of p' > b,
etc. is hard for me to swallow. Presumably there's an intermediate
stage so we'd have p' > b' > b. Yet no IE language that I'm aware of
preserves any trace of voiced implosives. Furthermore it's claimed
that the Glottalic theory accounts for the lack of *b because it's
interpretaed as p', the most marked version of p. But this is at odds
with *b/p' having a typical distribution non-initially.

There is another proposal that seems pretty obvious to me, yet I've
seen no discussion of it (though I can't imagine I'm the first to see
it). I would reinterpret the Traditional unvoiced stops as unvoiced
aspirates, the voiced stops and unvoiced stops, and the voiced
aspirates as plain voiced. Such a system would be typologically
acceptable (it's found in Ancient Greek). The Traditional unvoiced
stops often become either unvoiced aspirates or fricatives in
daughter languages which makes the reinterpretation plausable. The
Tradition voiced stops become unvoiced in Germanic and Armenian, as
well as some minor languages. This reinterpretation then would have
these languages as relic areas. Finally the Traditional voiced
aspirates become simply voiced in most daughters, which means that in
my reinterpretation the voiced stops remain unchanged in most
daughter languages.

This reinterpretation thus has two advantages:
1) It's typologically natural.
2) It stays fairly close to the comparative data.