Why? PLA can definitely afford to replace a mere 600 J7 with J10 around 10 years. As a fraction of GDP, J-10 is probably cheaper than J7 in the 80s. Again, PLA now spend more per soldier than Taiwan -- money is not a major issue.
Yes, but back then, the entire PLAAF's fighter budget was pretty much all going towards buying and maintaining J7s, J8s since they were the only options available. Today, they have quite a few more other options, which don't exactly come cheap either.
In addition to affordability, there is also the issue of necessity.
Does the PLAAF really need to replace J7s on a one-for-one basis with J10s? Quite clearly that is a 'no'. The J10s are far more capable planes, with much greater range, weapons payload and super in any other performance indicator you could care to use. You simply do not need as many J10s as J7s to provide the same level of defense for the same amount of airspace.
China might have more money, but that is hardly reason to go wasting it. America is richer than China per head and in aggregate, but even they cannot afford to not think about costs (as hard as that may be to believe at times).
Also, lets not forget that even if the PLAAF is dead set in maintaining their current numbers and wants to replace J7s on a one-for-one basis, the J10 and JF17 are not the only options. The JL9 and L15 are also fine candidates as J7 replacements for the air policing role in low threat areas. The plus side is that since the PLAAF would need to shell out for a new AJT anyways, it is very appealing to spend a few $m more per plane to make them fully combat capable as well.
This will enhance their training usefulness, and also allow the PLAAF to just use them for peacetime low threat condition air policing and patroling durties and same on needing a J7 type altogther in their order of battle.
A dedicated CAS JF17 sounds interesting, but it would appear that the PLA cares little about CAS, or else there would have been a far more serious effort into developing a Q5 follow-up, which is what the PLA needs if they want a new CAS platform, since such platforms are best designed as such from the ground up. For CAS, a JF17 is little different from a J10, so there is little incentive in coming up with a new J17 variant to do something the J10 (and JF17 itself) should already be able to do just as well (unless you go for a radical re-design, but in that case, you are probably better off starting from scratch with a blank slate anyways).