I wouldn't worry about it. In both cases, the numbers have no deeper meaning other than that they assign numeric values to the verbal response categories of an ordinal scale. In one case, the ordinal scale has 5, in the other 7 values. This will not "distort" anything.

Christian Geiser's answer is correct IF your analysis treats the data as ordinal. Since many people analyze Likert scales like they are interval data, you very well might get distortion if you did this. Use a nonparametric analysis and you should be fine!

Unless you are modifying a standardised scale, it should't matter.

You have two options to do: 1) using another software designed in terms of Item-Response Theory model, such as IRTpro 3, 2) using SPSS by Menu > Transform > Recode into same variable.... The second is easy to perform: so you can reduce 7 points into 5 points scale, but this cannot be happen inversely, i.e., changing 5 points into 7 points scale. In addition to this, I strongly agree with professor McLean, you must treat Likert type data as ordinal and do data analysis by means of non-parametric tests. And you might get distortion if you did any method other than this.