A recommendation is that you can perform an exploratory analysis to determine the structure of your model as it can be grouped into more or less factors than proposed, you can also take into account the covariance of errors.

Another thing that can improve your study is to use a robust estimator that includes other programs such as R, JASP or MPLUS because the masters use the maximum likelihood.

I actually am not sure that I want to start this, but many of us would argue that if the CFA theoretical model (path diagram of latents and indicators, etc.) comes from an EFA, then the CFA on the same data is no longer a CFA; it is no longer confirmatory but exploratory. More data would need to be collected after the EFA, then a CFA with the path model adopted from the EFA on those new data would actually be confirmatory. I realize this is a tough position in respect to practice, but, it is nonetheless my position.

Of course the same is true in "removing items" to improve fit indices. I might also say the the chi-squared simply can not be ignored. It is fine to use modification indices, intuition, divine revelation, or whatever sources one can muster to change the model to achieve better fit, and lack of disconfirmation (chi-squared), however, this is no longer a confirmatory analysis.

Dear David Adebesin; Hopefully those articles help you to fix your problems.