Are you asking about the results a measurement model within PLS-SEM? If so, why would you discard a variable that has a strong affiliation with a latent variable? If it's the case that this variable is the only one with any non-ignorable association with the latent variable, then the relevant concern is that you really don't have a factor per se, you merely have an individual variable whose variation is not shared with others and is apparently mostly specific/unique. So-called "singleton" factors are often dropped as unproductive.
If you're asking about something completely different, perhaps if you were to elaborate your query somewhat about what variables, what model, and what analytic scheme you were attempting, you'd get more helpful answers.
@Hina M. Hanif, your question is not clear. You need to be specific. But still, variable/parameter loading of 0.953 is very significant whether you run PCA, FA, PLS, etc.
According to Hair et al. (2019); Ringle et al. (2018); Sarstedt et al. (2017) the factor loadings between .5 and .65 can be retained if the CR & AVE are greater than 0.5 while factor loadings greater than 0.85 denote to CMB/CMB which can be due to SDE, SmartPLS (https://www.youtube.com/watch?v=J7eeu4O80_M) caters CMB/CMV with 'innver VIF matrix' if you see any number greater than 3.3 your data has SDE (Kock, 2015). SmartPLS has no remedy for CMV/CMB.
Also, certain self reported bias (https://www.youtube.com/watch?v=G0LNUwZnPqg) conclude on loadings and CB alpha greater than .85 as well. Run Harman's single-factor test (https://www.youtube.com/watch?v=xtet5_inRXk) on your data set in IBM SPSS and check the total variance explained for 1 factor, if it is below 50% you can retain your factors with loading up to 0.95--if it is greater than 50% then collect your data again using a questionnaire with different Likert scale for each variable. (https://www.youtube.com/watch?v=VSPcaHrBxa0)
Factor loading greater than .085 may also be a result of dichotomous/ordinal scale.