I made a comment on this seq at the end of 2013.  Looking at my comment and an earlier one by Juri-Stepan Gerasimov, we see the following:

The seq is primes adjacent to a number with at most 2 prime factors.  An example is 7, adjacent to 6&8.  The factors of 6 are 2,3  and   2,4.

Primes which do NOT solve a+b = a^2-b^2, with consecutive composites a and b with a > b

Primes which not are the sum of two consecutive composite numbers.

We can thus equate these as:

the sum of two consecutive composite numbers sieve the same primes as the difference of those same composites' squares.

So, A079149 includes ...23, 37, 47,...

(a,b)=8,9 gives 17

(a,b)=14,15 gives 29

(a,b)=15,16 gives 31

(a,b)=20,21 gives 41

We can thus xpress 17 as 9^2-8^2,  41 as 21^2-20^2.  We CANNOT xpress 23 or 37 similarly from two composites (C1,C2)

So, we know  A079149  primes  ! = C2^2-C1^2  whereas all other primes are.  Since there are 46 elements<=1000, and primepi(1000)=168, the majority of primes thus CAN be xpressed as the difference of such squares (46/168 ~27% ).  At 10K, we have 224/1229 ~ 18%